@livekit/agents 0.6.3 → 0.7.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (185) hide show
  1. package/dist/index.cjs +6 -1
  2. package/dist/index.cjs.map +1 -1
  3. package/dist/index.d.ts +3 -1
  4. package/dist/index.d.ts.map +1 -1
  5. package/dist/index.js +3 -0
  6. package/dist/index.js.map +1 -1
  7. package/dist/inference_runner.cjs +38 -0
  8. package/dist/inference_runner.cjs.map +1 -0
  9. package/dist/inference_runner.d.ts +11 -0
  10. package/dist/inference_runner.d.ts.map +1 -0
  11. package/dist/inference_runner.js +14 -0
  12. package/dist/inference_runner.js.map +1 -0
  13. package/dist/ipc/index.cjs +23 -0
  14. package/dist/ipc/index.cjs.map +1 -0
  15. package/dist/ipc/index.d.ts +2 -0
  16. package/dist/ipc/index.d.ts.map +1 -0
  17. package/dist/ipc/index.js +2 -0
  18. package/dist/ipc/index.js.map +1 -0
  19. package/dist/ipc/inference_executor.cjs +17 -0
  20. package/dist/ipc/inference_executor.cjs.map +1 -0
  21. package/dist/ipc/inference_executor.d.ts +4 -0
  22. package/dist/ipc/inference_executor.d.ts.map +1 -0
  23. package/dist/ipc/inference_executor.js +1 -0
  24. package/dist/ipc/inference_executor.js.map +1 -0
  25. package/dist/ipc/inference_proc_executor.cjs +97 -0
  26. package/dist/ipc/inference_proc_executor.cjs.map +1 -0
  27. package/dist/ipc/inference_proc_executor.d.ts +23 -0
  28. package/dist/ipc/inference_proc_executor.d.ts.map +1 -0
  29. package/dist/ipc/inference_proc_executor.js +72 -0
  30. package/dist/ipc/inference_proc_executor.js.map +1 -0
  31. package/dist/ipc/inference_proc_lazy_main.cjs +90 -0
  32. package/dist/ipc/inference_proc_lazy_main.cjs.map +1 -0
  33. package/dist/ipc/inference_proc_lazy_main.d.ts +2 -0
  34. package/dist/ipc/inference_proc_lazy_main.d.ts.map +1 -0
  35. package/dist/ipc/inference_proc_lazy_main.js +67 -0
  36. package/dist/ipc/inference_proc_lazy_main.js.map +1 -0
  37. package/dist/ipc/job_executor.cjs +8 -7
  38. package/dist/ipc/job_executor.cjs.map +1 -1
  39. package/dist/ipc/job_executor.d.ts +14 -15
  40. package/dist/ipc/job_executor.d.ts.map +1 -1
  41. package/dist/ipc/job_executor.js +7 -6
  42. package/dist/ipc/job_executor.js.map +1 -1
  43. package/dist/ipc/job_proc_executor.cjs +108 -0
  44. package/dist/ipc/job_proc_executor.cjs.map +1 -0
  45. package/dist/ipc/job_proc_executor.d.ts +19 -0
  46. package/dist/ipc/job_proc_executor.d.ts.map +1 -0
  47. package/dist/ipc/job_proc_executor.js +83 -0
  48. package/dist/ipc/job_proc_executor.js.map +1 -0
  49. package/dist/ipc/{job_main.cjs → job_proc_lazy_main.cjs} +41 -36
  50. package/dist/ipc/job_proc_lazy_main.cjs.map +1 -0
  51. package/dist/ipc/job_proc_lazy_main.d.ts +2 -0
  52. package/dist/ipc/job_proc_lazy_main.d.ts.map +1 -0
  53. package/dist/ipc/{job_main.js → job_proc_lazy_main.js} +41 -11
  54. package/dist/ipc/job_proc_lazy_main.js.map +1 -0
  55. package/dist/ipc/message.cjs.map +1 -1
  56. package/dist/ipc/message.d.ts +17 -0
  57. package/dist/ipc/message.d.ts.map +1 -1
  58. package/dist/ipc/proc_pool.cjs +30 -4
  59. package/dist/ipc/proc_pool.cjs.map +1 -1
  60. package/dist/ipc/proc_pool.d.ts +5 -1
  61. package/dist/ipc/proc_pool.d.ts.map +1 -1
  62. package/dist/ipc/proc_pool.js +30 -4
  63. package/dist/ipc/proc_pool.js.map +1 -1
  64. package/dist/ipc/{proc_job_executor.cjs → supervised_proc.cjs} +58 -46
  65. package/dist/ipc/supervised_proc.cjs.map +1 -0
  66. package/dist/ipc/supervised_proc.d.ts +30 -0
  67. package/dist/ipc/supervised_proc.d.ts.map +1 -0
  68. package/dist/ipc/{proc_job_executor.js → supervised_proc.js} +54 -32
  69. package/dist/ipc/supervised_proc.js.map +1 -0
  70. package/dist/job.cjs +18 -1
  71. package/dist/job.cjs.map +1 -1
  72. package/dist/job.d.ts +9 -1
  73. package/dist/job.d.ts.map +1 -1
  74. package/dist/job.js +17 -1
  75. package/dist/job.js.map +1 -1
  76. package/dist/metrics/base.cjs +2 -2
  77. package/dist/metrics/base.cjs.map +1 -1
  78. package/dist/metrics/base.d.ts +1 -1
  79. package/dist/metrics/base.d.ts.map +1 -1
  80. package/dist/metrics/base.js +2 -2
  81. package/dist/metrics/base.js.map +1 -1
  82. package/dist/multimodal/agent_playout.cjs +13 -14
  83. package/dist/multimodal/agent_playout.cjs.map +1 -1
  84. package/dist/multimodal/agent_playout.d.ts +4 -4
  85. package/dist/multimodal/agent_playout.d.ts.map +1 -1
  86. package/dist/multimodal/agent_playout.js +13 -14
  87. package/dist/multimodal/agent_playout.js.map +1 -1
  88. package/dist/multimodal/multimodal_agent.cjs +12 -8
  89. package/dist/multimodal/multimodal_agent.cjs.map +1 -1
  90. package/dist/multimodal/multimodal_agent.d.ts.map +1 -1
  91. package/dist/multimodal/multimodal_agent.js +13 -9
  92. package/dist/multimodal/multimodal_agent.js.map +1 -1
  93. package/dist/pipeline/agent_output.cjs +20 -4
  94. package/dist/pipeline/agent_output.cjs.map +1 -1
  95. package/dist/pipeline/agent_output.d.ts +4 -2
  96. package/dist/pipeline/agent_output.d.ts.map +1 -1
  97. package/dist/pipeline/agent_output.js +20 -4
  98. package/dist/pipeline/agent_output.js.map +1 -1
  99. package/dist/pipeline/agent_playout.cjs +9 -3
  100. package/dist/pipeline/agent_playout.cjs.map +1 -1
  101. package/dist/pipeline/agent_playout.d.ts +4 -2
  102. package/dist/pipeline/agent_playout.d.ts.map +1 -1
  103. package/dist/pipeline/agent_playout.js +9 -3
  104. package/dist/pipeline/agent_playout.js.map +1 -1
  105. package/dist/pipeline/human_input.cjs +6 -0
  106. package/dist/pipeline/human_input.cjs.map +1 -1
  107. package/dist/pipeline/human_input.d.ts +3 -1
  108. package/dist/pipeline/human_input.d.ts.map +1 -1
  109. package/dist/pipeline/human_input.js +6 -0
  110. package/dist/pipeline/human_input.js.map +1 -1
  111. package/dist/pipeline/pipeline_agent.cjs +79 -12
  112. package/dist/pipeline/pipeline_agent.cjs.map +1 -1
  113. package/dist/pipeline/pipeline_agent.d.ts +8 -0
  114. package/dist/pipeline/pipeline_agent.d.ts.map +1 -1
  115. package/dist/pipeline/pipeline_agent.js +79 -12
  116. package/dist/pipeline/pipeline_agent.js.map +1 -1
  117. package/dist/stt/stream_adapter.cjs +16 -4
  118. package/dist/stt/stream_adapter.cjs.map +1 -1
  119. package/dist/stt/stream_adapter.d.ts.map +1 -1
  120. package/dist/stt/stream_adapter.js +16 -4
  121. package/dist/stt/stream_adapter.js.map +1 -1
  122. package/dist/tokenize/basic/basic.cjs +2 -0
  123. package/dist/tokenize/basic/basic.cjs.map +1 -1
  124. package/dist/tokenize/basic/basic.d.ts +2 -0
  125. package/dist/tokenize/basic/basic.d.ts.map +1 -1
  126. package/dist/tokenize/basic/basic.js +1 -0
  127. package/dist/tokenize/basic/basic.js.map +1 -1
  128. package/dist/tokenize/basic/index.cjs +2 -0
  129. package/dist/tokenize/basic/index.cjs.map +1 -1
  130. package/dist/tokenize/basic/index.d.ts +1 -1
  131. package/dist/tokenize/basic/index.d.ts.map +1 -1
  132. package/dist/tokenize/basic/index.js +8 -1
  133. package/dist/tokenize/basic/index.js.map +1 -1
  134. package/dist/tokenize/token_stream.cjs +5 -3
  135. package/dist/tokenize/token_stream.cjs.map +1 -1
  136. package/dist/tokenize/token_stream.d.ts.map +1 -1
  137. package/dist/tokenize/token_stream.js +5 -3
  138. package/dist/tokenize/token_stream.js.map +1 -1
  139. package/dist/transcription.cjs +203 -86
  140. package/dist/transcription.cjs.map +1 -1
  141. package/dist/transcription.d.ts +24 -17
  142. package/dist/transcription.d.ts.map +1 -1
  143. package/dist/transcription.js +201 -85
  144. package/dist/transcription.js.map +1 -1
  145. package/dist/worker.cjs +42 -9
  146. package/dist/worker.cjs.map +1 -1
  147. package/dist/worker.d.ts +5 -1
  148. package/dist/worker.d.ts.map +1 -1
  149. package/dist/worker.js +42 -9
  150. package/dist/worker.js.map +1 -1
  151. package/package.json +3 -3
  152. package/src/index.ts +3 -1
  153. package/src/inference_runner.ts +19 -0
  154. package/src/ipc/index.ts +5 -0
  155. package/src/ipc/inference_executor.ts +7 -0
  156. package/src/ipc/inference_proc_executor.ts +93 -0
  157. package/src/ipc/inference_proc_lazy_main.ts +86 -0
  158. package/src/ipc/job_executor.ts +15 -17
  159. package/src/ipc/job_proc_executor.ts +112 -0
  160. package/src/ipc/{job_main.ts → job_proc_lazy_main.ts} +44 -14
  161. package/src/ipc/message.ts +14 -1
  162. package/src/ipc/proc_pool.ts +33 -3
  163. package/src/ipc/{proc_job_executor.ts → supervised_proc.ts} +80 -30
  164. package/src/job.ts +21 -0
  165. package/src/metrics/base.ts +7 -10
  166. package/src/multimodal/agent_playout.ts +14 -16
  167. package/src/multimodal/multimodal_agent.ts +13 -9
  168. package/src/pipeline/agent_output.ts +34 -5
  169. package/src/pipeline/agent_playout.ts +10 -1
  170. package/src/pipeline/human_input.ts +8 -0
  171. package/src/pipeline/pipeline_agent.ts +96 -11
  172. package/src/stt/stream_adapter.ts +17 -5
  173. package/src/tokenize/basic/basic.ts +2 -0
  174. package/src/tokenize/basic/index.ts +7 -1
  175. package/src/tokenize/token_stream.ts +6 -3
  176. package/src/transcription.ts +270 -96
  177. package/src/worker.ts +42 -5
  178. package/dist/ipc/job_main.cjs.map +0 -1
  179. package/dist/ipc/job_main.d.ts +0 -8
  180. package/dist/ipc/job_main.d.ts.map +0 -1
  181. package/dist/ipc/job_main.js.map +0 -1
  182. package/dist/ipc/proc_job_executor.cjs.map +0 -1
  183. package/dist/ipc/proc_job_executor.d.ts +0 -15
  184. package/dist/ipc/proc_job_executor.d.ts.map +0 -1
  185. package/dist/ipc/proc_job_executor.js.map +0 -1
@@ -33,7 +33,7 @@ class PlayoutHandle extends import_node_events.EventEmitter {
33
33
  #itemId;
34
34
  #contentIndex;
35
35
  /** @internal */
36
- transcriptionFwd;
36
+ synchronizer;
37
37
  /** @internal */
38
38
  doneFut;
39
39
  /** @internal */
@@ -45,13 +45,13 @@ class PlayoutHandle extends import_node_events.EventEmitter {
45
45
  /** @internal */
46
46
  totalPlayedTime;
47
47
  // Set when playout is done
48
- constructor(audioSource, sampleRate, itemId, contentIndex, transcriptionFwd) {
48
+ constructor(audioSource, sampleRate, itemId, contentIndex, synchronizer) {
49
49
  super();
50
50
  this.#audioSource = audioSource;
51
51
  this.#sampleRate = sampleRate;
52
52
  this.#itemId = itemId;
53
53
  this.#contentIndex = contentIndex;
54
- this.transcriptionFwd = transcriptionFwd;
54
+ this.synchronizer = synchronizer;
55
55
  this.doneFut = new import_utils.Future();
56
56
  this.intFut = new import_utils.Future();
57
57
  this.#interrupted = false;
@@ -70,7 +70,7 @@ class PlayoutHandle extends import_node_events.EventEmitter {
70
70
  );
71
71
  }
72
72
  get textChars() {
73
- return this.transcriptionFwd.currentCharacterIndex;
73
+ return this.synchronizer.playedText.length;
74
74
  }
75
75
  get contentIndex() {
76
76
  return this.#contentIndex;
@@ -103,13 +103,13 @@ class AgentPlayout extends import_node_events.EventEmitter {
103
103
  this.#inFrameSize = inFrameSize;
104
104
  this.#outFrameSize = outFrameSize;
105
105
  }
106
- play(itemId, contentIndex, transcriptionFwd, textStream, audioStream) {
106
+ play(itemId, contentIndex, synchronizer, textStream, audioStream) {
107
107
  const handle = new PlayoutHandle(
108
108
  this.#audioSource,
109
109
  this.#sampleRate,
110
110
  itemId,
111
111
  contentIndex,
112
- transcriptionFwd
112
+ synchronizer
113
113
  );
114
114
  this.#playoutTask = this.#makePlayoutTask(this.#playoutTask, handle, textStream, audioStream);
115
115
  return handle;
@@ -137,8 +137,9 @@ class AgentPlayout extends import_node_events.EventEmitter {
137
137
  if (cancelledText || cancelled) {
138
138
  break;
139
139
  }
140
- handle.transcriptionFwd.pushText(text);
140
+ handle.synchronizer.pushText(text);
141
141
  }
142
+ handle.synchronizer.markTextSegmentEnd();
142
143
  resolveText();
143
144
  } catch (error) {
144
145
  rejectText(error);
@@ -163,11 +164,11 @@ class AgentPlayout extends import_node_events.EventEmitter {
163
164
  break;
164
165
  }
165
166
  if (firstFrame) {
166
- handle.transcriptionFwd.start();
167
+ handle.synchronizer.segmentPlayoutStarted();
167
168
  this.emit("playout_started");
168
169
  firstFrame = false;
169
170
  }
170
- handle.transcriptionFwd.pushAudio(frame);
171
+ handle.synchronizer.pushAudio(frame);
171
172
  for (const f of bstream.write(frame.data.buffer)) {
172
173
  handle.pushedDuration += f.samplesPerChannel / f.sampleRate * 1e3;
173
174
  await this.#audioSource.captureFrame(f);
@@ -178,7 +179,7 @@ class AgentPlayout extends import_node_events.EventEmitter {
178
179
  handle.pushedDuration += f.samplesPerChannel / f.sampleRate * 1e3;
179
180
  await this.#audioSource.captureFrame(f);
180
181
  }
181
- handle.transcriptionFwd.markAudioComplete();
182
+ handle.synchronizer.markAudioSegmentEnd();
182
183
  await this.#audioSource.waitForPlayout();
183
184
  }
184
185
  resolveCapture();
@@ -197,19 +198,17 @@ class AgentPlayout extends import_node_events.EventEmitter {
197
198
  }
198
199
  handle.totalPlayedTime = handle.pushedDuration - this.#audioSource.queuedDuration;
199
200
  if (handle.interrupted || captureTask.error) {
201
+ await handle.synchronizer.close(true);
200
202
  this.#audioSource.clearQueue();
201
203
  }
202
204
  if (!readTextTask.isCancelled) {
203
205
  await (0, import_utils.gracefullyCancel)(readTextTask);
204
206
  }
205
207
  if (!firstFrame) {
206
- if (!handle.interrupted) {
207
- handle.transcriptionFwd.markTextComplete();
208
- }
209
208
  this.emit("playout_stopped", handle.interrupted);
210
209
  }
211
210
  handle.doneFut.resolve();
212
- await handle.transcriptionFwd.close(handle.interrupted);
211
+ await handle.synchronizer.close(false);
213
212
  }
214
213
  resolve();
215
214
  } catch (error) {
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/multimodal/agent_playout.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type AudioSource } from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport type { TranscriptionForwarder } from '../transcription.js';\nimport { type AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';\n\nexport const proto = {};\n\nexport class PlayoutHandle extends EventEmitter {\n #audioSource: AudioSource;\n #sampleRate: number;\n #itemId: string;\n #contentIndex: number;\n /** @internal */\n transcriptionFwd: TranscriptionForwarder;\n /** @internal */\n doneFut: Future;\n /** @internal */\n intFut: Future;\n /** @internal */\n #interrupted: boolean;\n /** @internal */\n pushedDuration: number;\n /** @internal */\n totalPlayedTime: number | undefined; // Set when playout is done\n\n constructor(\n audioSource: AudioSource,\n sampleRate: number,\n itemId: string,\n contentIndex: number,\n transcriptionFwd: TranscriptionForwarder,\n ) {\n super();\n this.#audioSource = audioSource;\n this.#sampleRate = sampleRate;\n this.#itemId = itemId;\n this.#contentIndex = contentIndex;\n this.transcriptionFwd = transcriptionFwd;\n this.doneFut = new Future();\n this.intFut = new Future();\n this.#interrupted = false;\n this.pushedDuration = 0;\n this.totalPlayedTime = undefined;\n }\n\n get itemId(): string {\n return this.#itemId;\n }\n\n get audioSamples(): number {\n if (this.totalPlayedTime !== undefined) {\n return Math.floor(this.totalPlayedTime * this.#sampleRate);\n }\n\n return Math.floor(\n (this.pushedDuration - this.#audioSource.queuedDuration) * (this.#sampleRate / 1000),\n );\n }\n\n get textChars(): number {\n return this.transcriptionFwd.currentCharacterIndex;\n }\n\n get contentIndex(): number {\n return this.#contentIndex;\n }\n\n get interrupted(): boolean {\n return this.#interrupted;\n }\n\n get done(): boolean {\n return this.doneFut.done || this.#interrupted;\n }\n\n interrupt() {\n if (this.doneFut.done) return;\n this.intFut.resolve();\n this.#interrupted = true;\n }\n}\n\nexport class AgentPlayout extends EventEmitter {\n #audioSource: AudioSource;\n #playoutTask: CancellablePromise<void> | null;\n #sampleRate: number;\n #numChannels: number;\n #inFrameSize: number;\n #outFrameSize: number;\n constructor(\n audioSource: AudioSource,\n sampleRate: number,\n numChannels: number,\n inFrameSize: number,\n outFrameSize: number,\n ) {\n super();\n this.#audioSource = audioSource;\n this.#playoutTask = null;\n this.#sampleRate = sampleRate;\n this.#numChannels = numChannels;\n this.#inFrameSize = inFrameSize;\n this.#outFrameSize = outFrameSize;\n }\n\n play(\n itemId: string,\n contentIndex: number,\n transcriptionFwd: TranscriptionForwarder,\n textStream: AsyncIterableQueue<string>,\n audioStream: AsyncIterableQueue<AudioFrame>,\n ): PlayoutHandle {\n const handle = new PlayoutHandle(\n this.#audioSource,\n this.#sampleRate,\n itemId,\n contentIndex,\n transcriptionFwd,\n );\n this.#playoutTask = this.#makePlayoutTask(this.#playoutTask, handle, textStream, audioStream);\n return handle;\n }\n\n #makePlayoutTask(\n oldTask: CancellablePromise<void> | null,\n handle: PlayoutHandle,\n textStream: AsyncIterableQueue<string>,\n audioStream: AsyncIterableQueue<AudioFrame>,\n ): CancellablePromise<void> {\n return new CancellablePromise<void>((resolve, reject, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n (async () => {\n try {\n if (oldTask) {\n await gracefullyCancel(oldTask);\n }\n\n let firstFrame = true;\n\n const readText = () =>\n new CancellablePromise<void>((resolveText, rejectText, onCancelText) => {\n let cancelledText = false;\n onCancelText(() => {\n cancelledText = true;\n });\n\n (async () => {\n try {\n for await (const text of textStream) {\n if (cancelledText || cancelled) {\n break;\n }\n handle.transcriptionFwd.pushText(text);\n }\n resolveText();\n } catch (error) {\n rejectText(error);\n }\n })();\n });\n\n const capture = () =>\n new CancellablePromise<void>((resolveCapture, rejectCapture, onCancelCapture) => {\n let cancelledCapture = false;\n onCancelCapture(() => {\n cancelledCapture = true;\n });\n\n (async () => {\n try {\n const samplesPerChannel = this.#outFrameSize;\n const bstream = new AudioByteStream(\n this.#sampleRate,\n this.#numChannels,\n samplesPerChannel,\n );\n\n for await (const frame of audioStream) {\n if (cancelledCapture || cancelled) {\n break;\n }\n if (firstFrame) {\n handle.transcriptionFwd.start();\n this.emit('playout_started');\n firstFrame = false;\n }\n\n handle.transcriptionFwd.pushAudio(frame);\n\n for (const f of bstream.write(frame.data.buffer)) {\n handle.pushedDuration += (f.samplesPerChannel / f.sampleRate) * 1000;\n await this.#audioSource.captureFrame(f);\n }\n }\n\n if (!cancelledCapture && !cancelled) {\n for (const f of bstream.flush()) {\n handle.pushedDuration += (f.samplesPerChannel / f.sampleRate) * 1000;\n await this.#audioSource.captureFrame(f);\n }\n\n handle.transcriptionFwd.markAudioComplete();\n\n await this.#audioSource.waitForPlayout();\n }\n\n resolveCapture();\n } catch (error) {\n rejectCapture(error);\n }\n })();\n });\n\n const readTextTask = readText();\n const captureTask = capture();\n\n try {\n await Promise.race([captureTask, handle.intFut.await]);\n } finally {\n if (!captureTask.isCancelled) {\n await gracefullyCancel(captureTask);\n }\n\n handle.totalPlayedTime = handle.pushedDuration - this.#audioSource.queuedDuration;\n\n if (handle.interrupted || captureTask.error) {\n this.#audioSource.clearQueue(); // make sure to remove any queued frames\n }\n\n if (!readTextTask.isCancelled) {\n await gracefullyCancel(readTextTask);\n }\n\n if (!firstFrame) {\n if (!handle.interrupted) {\n handle.transcriptionFwd.markTextComplete();\n }\n\n this.emit('playout_stopped', handle.interrupted);\n }\n\n handle.doneFut.resolve();\n await handle.transcriptionFwd.close(handle.interrupted);\n }\n\n resolve();\n } catch (error) {\n reject(error);\n }\n })();\n });\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAKA,yBAA6B;AAC7B,mBAAgC;AAEhC,mBAAsF;AAE/E,MAAM,QAAQ,CAAC;AAEf,MAAM,sBAAsB,gCAAa;AAAA,EAC9C;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA,YACE,aACA,YACA,QACA,cACA,kBACA;AACA,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,UAAU;AACf,SAAK,gBAAgB;AACrB,SAAK,mBAAmB;AACxB,SAAK,UAAU,IAAI,oBAAO;AAC1B,SAAK,SAAS,IAAI,oBAAO;AACzB,SAAK,eAAe;AACpB,SAAK,iBAAiB;AACtB,SAAK,kBAAkB;AAAA,EACzB;AAAA,EAEA,IAAI,SAAiB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,QAAI,KAAK,oBAAoB,QAAW;AACtC,aAAO,KAAK,MAAM,KAAK,kBAAkB,KAAK,WAAW;AAAA,IAC3D;AAEA,WAAO,KAAK;AAAA,OACT,KAAK,iBAAiB,KAAK,aAAa,mBAAmB,KAAK,cAAc;AAAA,IACjF;AAAA,EACF;AAAA,EAEA,IAAI,YAAoB;AACtB,WAAO,KAAK,iBAAiB;AAAA,EAC/B;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAgB;AAClB,WAAO,KAAK,QAAQ,QAAQ,KAAK;AAAA,EACnC;AAAA,EAEA,YAAY;AACV,QAAI,KAAK,QAAQ,KAAM;AACvB,SAAK,OAAO,QAAQ;AACpB,SAAK,eAAe;AAAA,EACtB;AACF;AAEO,MAAM,qBAAqB,gCAAa;AAAA,EAC7C;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,YACE,aACA,YACA,aACA,aACA,cACA;AACA,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,KACE,QACA,cACA,kBACA,YACA,aACe;AACf,UAAM,SAAS,IAAI;AAAA,MACjB,KAAK;AAAA,MACL,KAAK;AAAA,MACL;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,SAAK,eAAe,KAAK,iBAAiB,KAAK,cAAc,QAAQ,YAAY,WAAW;AAC5F,WAAO;AAAA,EACT;AAAA,EAEA,iBACE,SACA,QACA,YACA,aAC0B;AAC1B,WAAO,IAAI,gCAAyB,CAAC,SAAS,QAAQ,aAAa;AACjE,UAAI,YAAY;AAChB,eAAS,MAAM;AACb,oBAAY;AAAA,MACd,CAAC;AAED,OAAC,YAAY;AACX,YAAI;AACF,cAAI,SAAS;AACX,sBAAM,+BAAiB,OAAO;AAAA,UAChC;AAEA,cAAI,aAAa;AAEjB,gBAAM,WAAW,MACf,IAAI,gCAAyB,CAAC,aAAa,YAAY,iBAAiB;AACtE,gBAAI,gBAAgB;AACpB,yBAAa,MAAM;AACjB,8BAAgB;AAAA,YAClB,CAAC;AAED,aAAC,YAAY;AACX,kBAAI;AACF,iCAAiB,QAAQ,YAAY;AACnC,sBAAI,iBAAiB,WAAW;AAC9B;AAAA,kBACF;AACA,yBAAO,iBAAiB,SAAS,IAAI;AAAA,gBACvC;AACA,4BAAY;AAAA,cACd,SAAS,OAAO;AACd,2BAAW,KAAK;AAAA,cAClB;AAAA,YACF,GAAG;AAAA,UACL,CAAC;AAEH,gBAAM,UAAU,MACd,IAAI,gCAAyB,CAAC,gBAAgB,eAAe,oBAAoB;AAC/E,gBAAI,mBAAmB;AACvB,4BAAgB,MAAM;AACpB,iCAAmB;AAAA,YACrB,CAAC;AAED,aAAC,YAAY;AACX,kBAAI;AACF,sBAAM,oBAAoB,KAAK;AAC/B,sBAAM,UAAU,IAAI;AAAA,kBAClB,KAAK;AAAA,kBACL,KAAK;AAAA,kBACL;AAAA,gBACF;AAEA,iCAAiB,SAAS,aAAa;AACrC,sBAAI,oBAAoB,WAAW;AACjC;AAAA,kBACF;AACA,sBAAI,YAAY;AACd,2BAAO,iBAAiB,MAAM;AAC9B,yBAAK,KAAK,iBAAiB;AAC3B,iCAAa;AAAA,kBACf;AAEA,yBAAO,iBAAiB,UAAU,KAAK;AAEvC,6BAAW,KAAK,QAAQ,MAAM,MAAM,KAAK,MAAM,GAAG;AAChD,2BAAO,kBAAmB,EAAE,oBAAoB,EAAE,aAAc;AAChE,0BAAM,KAAK,aAAa,aAAa,CAAC;AAAA,kBACxC;AAAA,gBACF;AAEA,oBAAI,CAAC,oBAAoB,CAAC,WAAW;AACnC,6BAAW,KAAK,QAAQ,MAAM,GAAG;AAC/B,2BAAO,kBAAmB,EAAE,oBAAoB,EAAE,aAAc;AAChE,0BAAM,KAAK,aAAa,aAAa,CAAC;AAAA,kBACxC;AAEA,yBAAO,iBAAiB,kBAAkB;AAE1C,wBAAM,KAAK,aAAa,eAAe;AAAA,gBACzC;AAEA,+BAAe;AAAA,cACjB,SAAS,OAAO;AACd,8BAAc,KAAK;AAAA,cACrB;AAAA,YACF,GAAG;AAAA,UACL,CAAC;AAEH,gBAAM,eAAe,SAAS;AAC9B,gBAAM,cAAc,QAAQ;AAE5B,cAAI;AACF,kBAAM,QAAQ,KAAK,CAAC,aAAa,OAAO,OAAO,KAAK,CAAC;AAAA,UACvD,UAAE;AACA,gBAAI,CAAC,YAAY,aAAa;AAC5B,wBAAM,+BAAiB,WAAW;AAAA,YACpC;AAEA,mBAAO,kBAAkB,OAAO,iBAAiB,KAAK,aAAa;AAEnE,gBAAI,OAAO,eAAe,YAAY,OAAO;AAC3C,mBAAK,aAAa,WAAW;AAAA,YAC/B;AAEA,gBAAI,CAAC,aAAa,aAAa;AAC7B,wBAAM,+BAAiB,YAAY;AAAA,YACrC;AAEA,gBAAI,CAAC,YAAY;AACf,kBAAI,CAAC,OAAO,aAAa;AACvB,uBAAO,iBAAiB,iBAAiB;AAAA,cAC3C;AAEA,mBAAK,KAAK,mBAAmB,OAAO,WAAW;AAAA,YACjD;AAEA,mBAAO,QAAQ,QAAQ;AACvB,kBAAM,OAAO,iBAAiB,MAAM,OAAO,WAAW;AAAA,UACxD;AAEA,kBAAQ;AAAA,QACV,SAAS,OAAO;AACd,iBAAO,KAAK;AAAA,QACd;AAAA,MACF,GAAG;AAAA,IACL,CAAC;AAAA,EACH;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/multimodal/agent_playout.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type AudioSource } from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport type { TextAudioSynchronizer } from '../transcription.js';\nimport { type AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';\n\nexport const proto = {};\n\nexport class PlayoutHandle extends EventEmitter {\n #audioSource: AudioSource;\n #sampleRate: number;\n #itemId: string;\n #contentIndex: number;\n /** @internal */\n synchronizer: TextAudioSynchronizer;\n /** @internal */\n doneFut: Future;\n /** @internal */\n intFut: Future;\n /** @internal */\n #interrupted: boolean;\n /** @internal */\n pushedDuration: number;\n /** @internal */\n totalPlayedTime: number | undefined; // Set when playout is done\n\n constructor(\n audioSource: AudioSource,\n sampleRate: number,\n itemId: string,\n contentIndex: number,\n synchronizer: TextAudioSynchronizer,\n ) {\n super();\n this.#audioSource = audioSource;\n this.#sampleRate = sampleRate;\n this.#itemId = itemId;\n this.#contentIndex = contentIndex;\n this.synchronizer = synchronizer;\n this.doneFut = new Future();\n this.intFut = new Future();\n this.#interrupted = false;\n this.pushedDuration = 0;\n this.totalPlayedTime = undefined;\n }\n\n get itemId(): string {\n return this.#itemId;\n }\n\n get audioSamples(): number {\n if (this.totalPlayedTime !== undefined) {\n return Math.floor(this.totalPlayedTime * this.#sampleRate);\n }\n\n return Math.floor(\n (this.pushedDuration - this.#audioSource.queuedDuration) * (this.#sampleRate / 1000),\n );\n }\n\n get textChars(): number {\n return this.synchronizer.playedText.length;\n }\n\n get contentIndex(): number {\n return this.#contentIndex;\n }\n\n get interrupted(): boolean {\n return this.#interrupted;\n }\n\n get done(): boolean {\n return this.doneFut.done || this.#interrupted;\n }\n\n interrupt() {\n if (this.doneFut.done) return;\n this.intFut.resolve();\n this.#interrupted = true;\n }\n}\n\nexport class AgentPlayout extends EventEmitter {\n #audioSource: AudioSource;\n #playoutTask: CancellablePromise<void> | null;\n #sampleRate: number;\n #numChannels: number;\n #inFrameSize: number;\n #outFrameSize: number;\n constructor(\n audioSource: AudioSource,\n sampleRate: number,\n numChannels: number,\n inFrameSize: number,\n outFrameSize: number,\n ) {\n super();\n this.#audioSource = audioSource;\n this.#playoutTask = null;\n this.#sampleRate = sampleRate;\n this.#numChannels = numChannels;\n this.#inFrameSize = inFrameSize;\n this.#outFrameSize = outFrameSize;\n }\n\n play(\n itemId: string,\n contentIndex: number,\n synchronizer: TextAudioSynchronizer,\n textStream: AsyncIterableQueue<string>,\n audioStream: AsyncIterableQueue<AudioFrame>,\n ): PlayoutHandle {\n const handle = new PlayoutHandle(\n this.#audioSource,\n this.#sampleRate,\n itemId,\n contentIndex,\n synchronizer,\n );\n this.#playoutTask = this.#makePlayoutTask(this.#playoutTask, handle, textStream, audioStream);\n return handle;\n }\n\n #makePlayoutTask(\n oldTask: CancellablePromise<void> | null,\n handle: PlayoutHandle,\n textStream: AsyncIterableQueue<string>,\n audioStream: AsyncIterableQueue<AudioFrame>,\n ): CancellablePromise<void> {\n return new CancellablePromise<void>((resolve, reject, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n (async () => {\n try {\n if (oldTask) {\n await gracefullyCancel(oldTask);\n }\n\n let firstFrame = true;\n\n const readText = () =>\n new CancellablePromise<void>((resolveText, rejectText, onCancelText) => {\n let cancelledText = false;\n onCancelText(() => {\n cancelledText = true;\n });\n\n (async () => {\n try {\n for await (const text of textStream) {\n if (cancelledText || cancelled) {\n break;\n }\n handle.synchronizer.pushText(text);\n }\n handle.synchronizer.markTextSegmentEnd();\n resolveText();\n } catch (error) {\n rejectText(error);\n }\n })();\n });\n\n const capture = () =>\n new CancellablePromise<void>((resolveCapture, rejectCapture, onCancelCapture) => {\n let cancelledCapture = false;\n onCancelCapture(() => {\n cancelledCapture = true;\n });\n\n (async () => {\n try {\n const samplesPerChannel = this.#outFrameSize;\n const bstream = new AudioByteStream(\n this.#sampleRate,\n this.#numChannels,\n samplesPerChannel,\n );\n\n for await (const frame of audioStream) {\n if (cancelledCapture || cancelled) {\n break;\n }\n if (firstFrame) {\n handle.synchronizer.segmentPlayoutStarted();\n this.emit('playout_started');\n firstFrame = false;\n }\n\n handle.synchronizer.pushAudio(frame);\n\n for (const f of bstream.write(frame.data.buffer)) {\n handle.pushedDuration += (f.samplesPerChannel / f.sampleRate) * 1000;\n await this.#audioSource.captureFrame(f);\n }\n }\n\n if (!cancelledCapture && !cancelled) {\n for (const f of bstream.flush()) {\n handle.pushedDuration += (f.samplesPerChannel / f.sampleRate) * 1000;\n await this.#audioSource.captureFrame(f);\n }\n\n handle.synchronizer.markAudioSegmentEnd();\n\n await this.#audioSource.waitForPlayout();\n }\n\n resolveCapture();\n } catch (error) {\n rejectCapture(error);\n }\n })();\n });\n\n const readTextTask = readText();\n const captureTask = capture();\n\n try {\n await Promise.race([captureTask, handle.intFut.await]);\n } finally {\n if (!captureTask.isCancelled) {\n await gracefullyCancel(captureTask);\n }\n\n handle.totalPlayedTime = handle.pushedDuration - this.#audioSource.queuedDuration;\n\n if (handle.interrupted || captureTask.error) {\n await handle.synchronizer.close(true);\n this.#audioSource.clearQueue(); // make sure to remove any queued frames\n }\n\n if (!readTextTask.isCancelled) {\n await gracefullyCancel(readTextTask);\n }\n\n if (!firstFrame) {\n this.emit('playout_stopped', handle.interrupted);\n }\n\n handle.doneFut.resolve();\n await handle.synchronizer.close(false);\n }\n\n resolve();\n } catch (error) {\n reject(error);\n }\n })();\n });\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAKA,yBAA6B;AAC7B,mBAAgC;AAEhC,mBAAsF;AAE/E,MAAM,QAAQ,CAAC;AAEf,MAAM,sBAAsB,gCAAa;AAAA,EAC9C;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA,YACE,aACA,YACA,QACA,cACA,cACA;AACA,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,UAAU;AACf,SAAK,gBAAgB;AACrB,SAAK,eAAe;AACpB,SAAK,UAAU,IAAI,oBAAO;AAC1B,SAAK,SAAS,IAAI,oBAAO;AACzB,SAAK,eAAe;AACpB,SAAK,iBAAiB;AACtB,SAAK,kBAAkB;AAAA,EACzB;AAAA,EAEA,IAAI,SAAiB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,QAAI,KAAK,oBAAoB,QAAW;AACtC,aAAO,KAAK,MAAM,KAAK,kBAAkB,KAAK,WAAW;AAAA,IAC3D;AAEA,WAAO,KAAK;AAAA,OACT,KAAK,iBAAiB,KAAK,aAAa,mBAAmB,KAAK,cAAc;AAAA,IACjF;AAAA,EACF;AAAA,EAEA,IAAI,YAAoB;AACtB,WAAO,KAAK,aAAa,WAAW;AAAA,EACtC;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAgB;AAClB,WAAO,KAAK,QAAQ,QAAQ,KAAK;AAAA,EACnC;AAAA,EAEA,YAAY;AACV,QAAI,KAAK,QAAQ,KAAM;AACvB,SAAK,OAAO,QAAQ;AACpB,SAAK,eAAe;AAAA,EACtB;AACF;AAEO,MAAM,qBAAqB,gCAAa;AAAA,EAC7C;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,YACE,aACA,YACA,aACA,aACA,cACA;AACA,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,KACE,QACA,cACA,cACA,YACA,aACe;AACf,UAAM,SAAS,IAAI;AAAA,MACjB,KAAK;AAAA,MACL,KAAK;AAAA,MACL;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,SAAK,eAAe,KAAK,iBAAiB,KAAK,cAAc,QAAQ,YAAY,WAAW;AAC5F,WAAO;AAAA,EACT;AAAA,EAEA,iBACE,SACA,QACA,YACA,aAC0B;AAC1B,WAAO,IAAI,gCAAyB,CAAC,SAAS,QAAQ,aAAa;AACjE,UAAI,YAAY;AAChB,eAAS,MAAM;AACb,oBAAY;AAAA,MACd,CAAC;AAED,OAAC,YAAY;AACX,YAAI;AACF,cAAI,SAAS;AACX,sBAAM,+BAAiB,OAAO;AAAA,UAChC;AAEA,cAAI,aAAa;AAEjB,gBAAM,WAAW,MACf,IAAI,gCAAyB,CAAC,aAAa,YAAY,iBAAiB;AACtE,gBAAI,gBAAgB;AACpB,yBAAa,MAAM;AACjB,8BAAgB;AAAA,YAClB,CAAC;AAED,aAAC,YAAY;AACX,kBAAI;AACF,iCAAiB,QAAQ,YAAY;AACnC,sBAAI,iBAAiB,WAAW;AAC9B;AAAA,kBACF;AACA,yBAAO,aAAa,SAAS,IAAI;AAAA,gBACnC;AACA,uBAAO,aAAa,mBAAmB;AACvC,4BAAY;AAAA,cACd,SAAS,OAAO;AACd,2BAAW,KAAK;AAAA,cAClB;AAAA,YACF,GAAG;AAAA,UACL,CAAC;AAEH,gBAAM,UAAU,MACd,IAAI,gCAAyB,CAAC,gBAAgB,eAAe,oBAAoB;AAC/E,gBAAI,mBAAmB;AACvB,4BAAgB,MAAM;AACpB,iCAAmB;AAAA,YACrB,CAAC;AAED,aAAC,YAAY;AACX,kBAAI;AACF,sBAAM,oBAAoB,KAAK;AAC/B,sBAAM,UAAU,IAAI;AAAA,kBAClB,KAAK;AAAA,kBACL,KAAK;AAAA,kBACL;AAAA,gBACF;AAEA,iCAAiB,SAAS,aAAa;AACrC,sBAAI,oBAAoB,WAAW;AACjC;AAAA,kBACF;AACA,sBAAI,YAAY;AACd,2BAAO,aAAa,sBAAsB;AAC1C,yBAAK,KAAK,iBAAiB;AAC3B,iCAAa;AAAA,kBACf;AAEA,yBAAO,aAAa,UAAU,KAAK;AAEnC,6BAAW,KAAK,QAAQ,MAAM,MAAM,KAAK,MAAM,GAAG;AAChD,2BAAO,kBAAmB,EAAE,oBAAoB,EAAE,aAAc;AAChE,0BAAM,KAAK,aAAa,aAAa,CAAC;AAAA,kBACxC;AAAA,gBACF;AAEA,oBAAI,CAAC,oBAAoB,CAAC,WAAW;AACnC,6BAAW,KAAK,QAAQ,MAAM,GAAG;AAC/B,2BAAO,kBAAmB,EAAE,oBAAoB,EAAE,aAAc;AAChE,0BAAM,KAAK,aAAa,aAAa,CAAC;AAAA,kBACxC;AAEA,yBAAO,aAAa,oBAAoB;AAExC,wBAAM,KAAK,aAAa,eAAe;AAAA,gBACzC;AAEA,+BAAe;AAAA,cACjB,SAAS,OAAO;AACd,8BAAc,KAAK;AAAA,cACrB;AAAA,YACF,GAAG;AAAA,UACL,CAAC;AAEH,gBAAM,eAAe,SAAS;AAC9B,gBAAM,cAAc,QAAQ;AAE5B,cAAI;AACF,kBAAM,QAAQ,KAAK,CAAC,aAAa,OAAO,OAAO,KAAK,CAAC;AAAA,UACvD,UAAE;AACA,gBAAI,CAAC,YAAY,aAAa;AAC5B,wBAAM,+BAAiB,WAAW;AAAA,YACpC;AAEA,mBAAO,kBAAkB,OAAO,iBAAiB,KAAK,aAAa;AAEnE,gBAAI,OAAO,eAAe,YAAY,OAAO;AAC3C,oBAAM,OAAO,aAAa,MAAM,IAAI;AACpC,mBAAK,aAAa,WAAW;AAAA,YAC/B;AAEA,gBAAI,CAAC,aAAa,aAAa;AAC7B,wBAAM,+BAAiB,YAAY;AAAA,YACrC;AAEA,gBAAI,CAAC,YAAY;AACf,mBAAK,KAAK,mBAAmB,OAAO,WAAW;AAAA,YACjD;AAEA,mBAAO,QAAQ,QAAQ;AACvB,kBAAM,OAAO,aAAa,MAAM,KAAK;AAAA,UACvC;AAEA,kBAAQ;AAAA,QACV,SAAS,OAAO;AACd,iBAAO,KAAK;AAAA,QACd;AAAA,MACF,GAAG;AAAA,IACL,CAAC;AAAA,EACH;AACF;","names":[]}
@@ -2,13 +2,13 @@
2
2
  import type { AudioFrame } from '@livekit/rtc-node';
3
3
  import { type AudioSource } from '@livekit/rtc-node';
4
4
  import { EventEmitter } from 'node:events';
5
- import type { TranscriptionForwarder } from '../transcription.js';
5
+ import type { TextAudioSynchronizer } from '../transcription.js';
6
6
  import { type AsyncIterableQueue, Future } from '../utils.js';
7
7
  export declare const proto: {};
8
8
  export declare class PlayoutHandle extends EventEmitter {
9
9
  #private;
10
10
  /** @internal */
11
- transcriptionFwd: TranscriptionForwarder;
11
+ synchronizer: TextAudioSynchronizer;
12
12
  /** @internal */
13
13
  doneFut: Future;
14
14
  /** @internal */
@@ -17,7 +17,7 @@ export declare class PlayoutHandle extends EventEmitter {
17
17
  pushedDuration: number;
18
18
  /** @internal */
19
19
  totalPlayedTime: number | undefined;
20
- constructor(audioSource: AudioSource, sampleRate: number, itemId: string, contentIndex: number, transcriptionFwd: TranscriptionForwarder);
20
+ constructor(audioSource: AudioSource, sampleRate: number, itemId: string, contentIndex: number, synchronizer: TextAudioSynchronizer);
21
21
  get itemId(): string;
22
22
  get audioSamples(): number;
23
23
  get textChars(): number;
@@ -29,6 +29,6 @@ export declare class PlayoutHandle extends EventEmitter {
29
29
  export declare class AgentPlayout extends EventEmitter {
30
30
  #private;
31
31
  constructor(audioSource: AudioSource, sampleRate: number, numChannels: number, inFrameSize: number, outFrameSize: number);
32
- play(itemId: string, contentIndex: number, transcriptionFwd: TranscriptionForwarder, textStream: AsyncIterableQueue<string>, audioStream: AsyncIterableQueue<AudioFrame>): PlayoutHandle;
32
+ play(itemId: string, contentIndex: number, synchronizer: TextAudioSynchronizer, textStream: AsyncIterableQueue<string>, audioStream: AsyncIterableQueue<AudioFrame>): PlayoutHandle;
33
33
  }
34
34
  //# sourceMappingURL=agent_playout.d.ts.map
@@ -1 +1 @@
1
- {"version":3,"file":"agent_playout.d.ts","sourceRoot":"","sources":["../../src/multimodal/agent_playout.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,EAAE,KAAK,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACrD,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAE3C,OAAO,KAAK,EAAE,sBAAsB,EAAE,MAAM,qBAAqB,CAAC;AAClE,OAAO,EAAE,KAAK,kBAAkB,EAAsB,MAAM,EAAoB,MAAM,aAAa,CAAC;AAEpG,eAAO,MAAM,KAAK,IAAK,CAAC;AAExB,qBAAa,aAAc,SAAQ,YAAY;;IAK7C,gBAAgB;IAChB,gBAAgB,EAAE,sBAAsB,CAAC;IACzC,gBAAgB;IAChB,OAAO,EAAE,MAAM,CAAC;IAChB,gBAAgB;IAChB,MAAM,EAAE,MAAM,CAAC;IAGf,gBAAgB;IAChB,cAAc,EAAE,MAAM,CAAC;IACvB,gBAAgB;IAChB,eAAe,EAAE,MAAM,GAAG,SAAS,CAAC;gBAGlC,WAAW,EAAE,WAAW,EACxB,UAAU,EAAE,MAAM,EAClB,MAAM,EAAE,MAAM,EACd,YAAY,EAAE,MAAM,EACpB,gBAAgB,EAAE,sBAAsB;IAe1C,IAAI,MAAM,IAAI,MAAM,CAEnB;IAED,IAAI,YAAY,IAAI,MAAM,CAQzB;IAED,IAAI,SAAS,IAAI,MAAM,CAEtB;IAED,IAAI,YAAY,IAAI,MAAM,CAEzB;IAED,IAAI,WAAW,IAAI,OAAO,CAEzB;IAED,IAAI,IAAI,IAAI,OAAO,CAElB;IAED,SAAS;CAKV;AAED,qBAAa,YAAa,SAAQ,YAAY;;gBAQ1C,WAAW,EAAE,WAAW,EACxB,UAAU,EAAE,MAAM,EAClB,WAAW,EAAE,MAAM,EACnB,WAAW,EAAE,MAAM,EACnB,YAAY,EAAE,MAAM;IAWtB,IAAI,CACF,MAAM,EAAE,MAAM,EACd,YAAY,EAAE,MAAM,EACpB,gBAAgB,EAAE,sBAAsB,EACxC,UAAU,EAAE,kBAAkB,CAAC,MAAM,CAAC,EACtC,WAAW,EAAE,kBAAkB,CAAC,UAAU,CAAC,GAC1C,aAAa;CAiJjB"}
1
+ {"version":3,"file":"agent_playout.d.ts","sourceRoot":"","sources":["../../src/multimodal/agent_playout.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,EAAE,KAAK,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACrD,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAE3C,OAAO,KAAK,EAAE,qBAAqB,EAAE,MAAM,qBAAqB,CAAC;AACjE,OAAO,EAAE,KAAK,kBAAkB,EAAsB,MAAM,EAAoB,MAAM,aAAa,CAAC;AAEpG,eAAO,MAAM,KAAK,IAAK,CAAC;AAExB,qBAAa,aAAc,SAAQ,YAAY;;IAK7C,gBAAgB;IAChB,YAAY,EAAE,qBAAqB,CAAC;IACpC,gBAAgB;IAChB,OAAO,EAAE,MAAM,CAAC;IAChB,gBAAgB;IAChB,MAAM,EAAE,MAAM,CAAC;IAGf,gBAAgB;IAChB,cAAc,EAAE,MAAM,CAAC;IACvB,gBAAgB;IAChB,eAAe,EAAE,MAAM,GAAG,SAAS,CAAC;gBAGlC,WAAW,EAAE,WAAW,EACxB,UAAU,EAAE,MAAM,EAClB,MAAM,EAAE,MAAM,EACd,YAAY,EAAE,MAAM,EACpB,YAAY,EAAE,qBAAqB;IAerC,IAAI,MAAM,IAAI,MAAM,CAEnB;IAED,IAAI,YAAY,IAAI,MAAM,CAQzB;IAED,IAAI,SAAS,IAAI,MAAM,CAEtB;IAED,IAAI,YAAY,IAAI,MAAM,CAEzB;IAED,IAAI,WAAW,IAAI,OAAO,CAEzB;IAED,IAAI,IAAI,IAAI,OAAO,CAElB;IAED,SAAS;CAKV;AAED,qBAAa,YAAa,SAAQ,YAAY;;gBAQ1C,WAAW,EAAE,WAAW,EACxB,UAAU,EAAE,MAAM,EAClB,WAAW,EAAE,MAAM,EACnB,WAAW,EAAE,MAAM,EACnB,YAAY,EAAE,MAAM;IAWtB,IAAI,CACF,MAAM,EAAE,MAAM,EACd,YAAY,EAAE,MAAM,EACpB,YAAY,EAAE,qBAAqB,EACnC,UAAU,EAAE,kBAAkB,CAAC,MAAM,CAAC,EACtC,WAAW,EAAE,kBAAkB,CAAC,UAAU,CAAC,GAC1C,aAAa;CA+IjB"}
@@ -8,7 +8,7 @@ class PlayoutHandle extends EventEmitter {
8
8
  #itemId;
9
9
  #contentIndex;
10
10
  /** @internal */
11
- transcriptionFwd;
11
+ synchronizer;
12
12
  /** @internal */
13
13
  doneFut;
14
14
  /** @internal */
@@ -20,13 +20,13 @@ class PlayoutHandle extends EventEmitter {
20
20
  /** @internal */
21
21
  totalPlayedTime;
22
22
  // Set when playout is done
23
- constructor(audioSource, sampleRate, itemId, contentIndex, transcriptionFwd) {
23
+ constructor(audioSource, sampleRate, itemId, contentIndex, synchronizer) {
24
24
  super();
25
25
  this.#audioSource = audioSource;
26
26
  this.#sampleRate = sampleRate;
27
27
  this.#itemId = itemId;
28
28
  this.#contentIndex = contentIndex;
29
- this.transcriptionFwd = transcriptionFwd;
29
+ this.synchronizer = synchronizer;
30
30
  this.doneFut = new Future();
31
31
  this.intFut = new Future();
32
32
  this.#interrupted = false;
@@ -45,7 +45,7 @@ class PlayoutHandle extends EventEmitter {
45
45
  );
46
46
  }
47
47
  get textChars() {
48
- return this.transcriptionFwd.currentCharacterIndex;
48
+ return this.synchronizer.playedText.length;
49
49
  }
50
50
  get contentIndex() {
51
51
  return this.#contentIndex;
@@ -78,13 +78,13 @@ class AgentPlayout extends EventEmitter {
78
78
  this.#inFrameSize = inFrameSize;
79
79
  this.#outFrameSize = outFrameSize;
80
80
  }
81
- play(itemId, contentIndex, transcriptionFwd, textStream, audioStream) {
81
+ play(itemId, contentIndex, synchronizer, textStream, audioStream) {
82
82
  const handle = new PlayoutHandle(
83
83
  this.#audioSource,
84
84
  this.#sampleRate,
85
85
  itemId,
86
86
  contentIndex,
87
- transcriptionFwd
87
+ synchronizer
88
88
  );
89
89
  this.#playoutTask = this.#makePlayoutTask(this.#playoutTask, handle, textStream, audioStream);
90
90
  return handle;
@@ -112,8 +112,9 @@ class AgentPlayout extends EventEmitter {
112
112
  if (cancelledText || cancelled) {
113
113
  break;
114
114
  }
115
- handle.transcriptionFwd.pushText(text);
115
+ handle.synchronizer.pushText(text);
116
116
  }
117
+ handle.synchronizer.markTextSegmentEnd();
117
118
  resolveText();
118
119
  } catch (error) {
119
120
  rejectText(error);
@@ -138,11 +139,11 @@ class AgentPlayout extends EventEmitter {
138
139
  break;
139
140
  }
140
141
  if (firstFrame) {
141
- handle.transcriptionFwd.start();
142
+ handle.synchronizer.segmentPlayoutStarted();
142
143
  this.emit("playout_started");
143
144
  firstFrame = false;
144
145
  }
145
- handle.transcriptionFwd.pushAudio(frame);
146
+ handle.synchronizer.pushAudio(frame);
146
147
  for (const f of bstream.write(frame.data.buffer)) {
147
148
  handle.pushedDuration += f.samplesPerChannel / f.sampleRate * 1e3;
148
149
  await this.#audioSource.captureFrame(f);
@@ -153,7 +154,7 @@ class AgentPlayout extends EventEmitter {
153
154
  handle.pushedDuration += f.samplesPerChannel / f.sampleRate * 1e3;
154
155
  await this.#audioSource.captureFrame(f);
155
156
  }
156
- handle.transcriptionFwd.markAudioComplete();
157
+ handle.synchronizer.markAudioSegmentEnd();
157
158
  await this.#audioSource.waitForPlayout();
158
159
  }
159
160
  resolveCapture();
@@ -172,19 +173,17 @@ class AgentPlayout extends EventEmitter {
172
173
  }
173
174
  handle.totalPlayedTime = handle.pushedDuration - this.#audioSource.queuedDuration;
174
175
  if (handle.interrupted || captureTask.error) {
176
+ await handle.synchronizer.close(true);
175
177
  this.#audioSource.clearQueue();
176
178
  }
177
179
  if (!readTextTask.isCancelled) {
178
180
  await gracefullyCancel(readTextTask);
179
181
  }
180
182
  if (!firstFrame) {
181
- if (!handle.interrupted) {
182
- handle.transcriptionFwd.markTextComplete();
183
- }
184
183
  this.emit("playout_stopped", handle.interrupted);
185
184
  }
186
185
  handle.doneFut.resolve();
187
- await handle.transcriptionFwd.close(handle.interrupted);
186
+ await handle.synchronizer.close(false);
188
187
  }
189
188
  resolve();
190
189
  } catch (error) {
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/multimodal/agent_playout.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type AudioSource } from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport type { TranscriptionForwarder } from '../transcription.js';\nimport { type AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';\n\nexport const proto = {};\n\nexport class PlayoutHandle extends EventEmitter {\n #audioSource: AudioSource;\n #sampleRate: number;\n #itemId: string;\n #contentIndex: number;\n /** @internal */\n transcriptionFwd: TranscriptionForwarder;\n /** @internal */\n doneFut: Future;\n /** @internal */\n intFut: Future;\n /** @internal */\n #interrupted: boolean;\n /** @internal */\n pushedDuration: number;\n /** @internal */\n totalPlayedTime: number | undefined; // Set when playout is done\n\n constructor(\n audioSource: AudioSource,\n sampleRate: number,\n itemId: string,\n contentIndex: number,\n transcriptionFwd: TranscriptionForwarder,\n ) {\n super();\n this.#audioSource = audioSource;\n this.#sampleRate = sampleRate;\n this.#itemId = itemId;\n this.#contentIndex = contentIndex;\n this.transcriptionFwd = transcriptionFwd;\n this.doneFut = new Future();\n this.intFut = new Future();\n this.#interrupted = false;\n this.pushedDuration = 0;\n this.totalPlayedTime = undefined;\n }\n\n get itemId(): string {\n return this.#itemId;\n }\n\n get audioSamples(): number {\n if (this.totalPlayedTime !== undefined) {\n return Math.floor(this.totalPlayedTime * this.#sampleRate);\n }\n\n return Math.floor(\n (this.pushedDuration - this.#audioSource.queuedDuration) * (this.#sampleRate / 1000),\n );\n }\n\n get textChars(): number {\n return this.transcriptionFwd.currentCharacterIndex;\n }\n\n get contentIndex(): number {\n return this.#contentIndex;\n }\n\n get interrupted(): boolean {\n return this.#interrupted;\n }\n\n get done(): boolean {\n return this.doneFut.done || this.#interrupted;\n }\n\n interrupt() {\n if (this.doneFut.done) return;\n this.intFut.resolve();\n this.#interrupted = true;\n }\n}\n\nexport class AgentPlayout extends EventEmitter {\n #audioSource: AudioSource;\n #playoutTask: CancellablePromise<void> | null;\n #sampleRate: number;\n #numChannels: number;\n #inFrameSize: number;\n #outFrameSize: number;\n constructor(\n audioSource: AudioSource,\n sampleRate: number,\n numChannels: number,\n inFrameSize: number,\n outFrameSize: number,\n ) {\n super();\n this.#audioSource = audioSource;\n this.#playoutTask = null;\n this.#sampleRate = sampleRate;\n this.#numChannels = numChannels;\n this.#inFrameSize = inFrameSize;\n this.#outFrameSize = outFrameSize;\n }\n\n play(\n itemId: string,\n contentIndex: number,\n transcriptionFwd: TranscriptionForwarder,\n textStream: AsyncIterableQueue<string>,\n audioStream: AsyncIterableQueue<AudioFrame>,\n ): PlayoutHandle {\n const handle = new PlayoutHandle(\n this.#audioSource,\n this.#sampleRate,\n itemId,\n contentIndex,\n transcriptionFwd,\n );\n this.#playoutTask = this.#makePlayoutTask(this.#playoutTask, handle, textStream, audioStream);\n return handle;\n }\n\n #makePlayoutTask(\n oldTask: CancellablePromise<void> | null,\n handle: PlayoutHandle,\n textStream: AsyncIterableQueue<string>,\n audioStream: AsyncIterableQueue<AudioFrame>,\n ): CancellablePromise<void> {\n return new CancellablePromise<void>((resolve, reject, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n (async () => {\n try {\n if (oldTask) {\n await gracefullyCancel(oldTask);\n }\n\n let firstFrame = true;\n\n const readText = () =>\n new CancellablePromise<void>((resolveText, rejectText, onCancelText) => {\n let cancelledText = false;\n onCancelText(() => {\n cancelledText = true;\n });\n\n (async () => {\n try {\n for await (const text of textStream) {\n if (cancelledText || cancelled) {\n break;\n }\n handle.transcriptionFwd.pushText(text);\n }\n resolveText();\n } catch (error) {\n rejectText(error);\n }\n })();\n });\n\n const capture = () =>\n new CancellablePromise<void>((resolveCapture, rejectCapture, onCancelCapture) => {\n let cancelledCapture = false;\n onCancelCapture(() => {\n cancelledCapture = true;\n });\n\n (async () => {\n try {\n const samplesPerChannel = this.#outFrameSize;\n const bstream = new AudioByteStream(\n this.#sampleRate,\n this.#numChannels,\n samplesPerChannel,\n );\n\n for await (const frame of audioStream) {\n if (cancelledCapture || cancelled) {\n break;\n }\n if (firstFrame) {\n handle.transcriptionFwd.start();\n this.emit('playout_started');\n firstFrame = false;\n }\n\n handle.transcriptionFwd.pushAudio(frame);\n\n for (const f of bstream.write(frame.data.buffer)) {\n handle.pushedDuration += (f.samplesPerChannel / f.sampleRate) * 1000;\n await this.#audioSource.captureFrame(f);\n }\n }\n\n if (!cancelledCapture && !cancelled) {\n for (const f of bstream.flush()) {\n handle.pushedDuration += (f.samplesPerChannel / f.sampleRate) * 1000;\n await this.#audioSource.captureFrame(f);\n }\n\n handle.transcriptionFwd.markAudioComplete();\n\n await this.#audioSource.waitForPlayout();\n }\n\n resolveCapture();\n } catch (error) {\n rejectCapture(error);\n }\n })();\n });\n\n const readTextTask = readText();\n const captureTask = capture();\n\n try {\n await Promise.race([captureTask, handle.intFut.await]);\n } finally {\n if (!captureTask.isCancelled) {\n await gracefullyCancel(captureTask);\n }\n\n handle.totalPlayedTime = handle.pushedDuration - this.#audioSource.queuedDuration;\n\n if (handle.interrupted || captureTask.error) {\n this.#audioSource.clearQueue(); // make sure to remove any queued frames\n }\n\n if (!readTextTask.isCancelled) {\n await gracefullyCancel(readTextTask);\n }\n\n if (!firstFrame) {\n if (!handle.interrupted) {\n handle.transcriptionFwd.markTextComplete();\n }\n\n this.emit('playout_stopped', handle.interrupted);\n }\n\n handle.doneFut.resolve();\n await handle.transcriptionFwd.close(handle.interrupted);\n }\n\n resolve();\n } catch (error) {\n reject(error);\n }\n })();\n });\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAC7B,SAAS,uBAAuB;AAEhC,SAAkC,oBAAoB,QAAQ,wBAAwB;AAE/E,MAAM,QAAQ,CAAC;AAEf,MAAM,sBAAsB,aAAa;AAAA,EAC9C;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA,YACE,aACA,YACA,QACA,cACA,kBACA;AACA,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,UAAU;AACf,SAAK,gBAAgB;AACrB,SAAK,mBAAmB;AACxB,SAAK,UAAU,IAAI,OAAO;AAC1B,SAAK,SAAS,IAAI,OAAO;AACzB,SAAK,eAAe;AACpB,SAAK,iBAAiB;AACtB,SAAK,kBAAkB;AAAA,EACzB;AAAA,EAEA,IAAI,SAAiB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,QAAI,KAAK,oBAAoB,QAAW;AACtC,aAAO,KAAK,MAAM,KAAK,kBAAkB,KAAK,WAAW;AAAA,IAC3D;AAEA,WAAO,KAAK;AAAA,OACT,KAAK,iBAAiB,KAAK,aAAa,mBAAmB,KAAK,cAAc;AAAA,IACjF;AAAA,EACF;AAAA,EAEA,IAAI,YAAoB;AACtB,WAAO,KAAK,iBAAiB;AAAA,EAC/B;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAgB;AAClB,WAAO,KAAK,QAAQ,QAAQ,KAAK;AAAA,EACnC;AAAA,EAEA,YAAY;AACV,QAAI,KAAK,QAAQ,KAAM;AACvB,SAAK,OAAO,QAAQ;AACpB,SAAK,eAAe;AAAA,EACtB;AACF;AAEO,MAAM,qBAAqB,aAAa;AAAA,EAC7C;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,YACE,aACA,YACA,aACA,aACA,cACA;AACA,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,KACE,QACA,cACA,kBACA,YACA,aACe;AACf,UAAM,SAAS,IAAI;AAAA,MACjB,KAAK;AAAA,MACL,KAAK;AAAA,MACL;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,SAAK,eAAe,KAAK,iBAAiB,KAAK,cAAc,QAAQ,YAAY,WAAW;AAC5F,WAAO;AAAA,EACT;AAAA,EAEA,iBACE,SACA,QACA,YACA,aAC0B;AAC1B,WAAO,IAAI,mBAAyB,CAAC,SAAS,QAAQ,aAAa;AACjE,UAAI,YAAY;AAChB,eAAS,MAAM;AACb,oBAAY;AAAA,MACd,CAAC;AAED,OAAC,YAAY;AACX,YAAI;AACF,cAAI,SAAS;AACX,kBAAM,iBAAiB,OAAO;AAAA,UAChC;AAEA,cAAI,aAAa;AAEjB,gBAAM,WAAW,MACf,IAAI,mBAAyB,CAAC,aAAa,YAAY,iBAAiB;AACtE,gBAAI,gBAAgB;AACpB,yBAAa,MAAM;AACjB,8BAAgB;AAAA,YAClB,CAAC;AAED,aAAC,YAAY;AACX,kBAAI;AACF,iCAAiB,QAAQ,YAAY;AACnC,sBAAI,iBAAiB,WAAW;AAC9B;AAAA,kBACF;AACA,yBAAO,iBAAiB,SAAS,IAAI;AAAA,gBACvC;AACA,4BAAY;AAAA,cACd,SAAS,OAAO;AACd,2BAAW,KAAK;AAAA,cAClB;AAAA,YACF,GAAG;AAAA,UACL,CAAC;AAEH,gBAAM,UAAU,MACd,IAAI,mBAAyB,CAAC,gBAAgB,eAAe,oBAAoB;AAC/E,gBAAI,mBAAmB;AACvB,4BAAgB,MAAM;AACpB,iCAAmB;AAAA,YACrB,CAAC;AAED,aAAC,YAAY;AACX,kBAAI;AACF,sBAAM,oBAAoB,KAAK;AAC/B,sBAAM,UAAU,IAAI;AAAA,kBAClB,KAAK;AAAA,kBACL,KAAK;AAAA,kBACL;AAAA,gBACF;AAEA,iCAAiB,SAAS,aAAa;AACrC,sBAAI,oBAAoB,WAAW;AACjC;AAAA,kBACF;AACA,sBAAI,YAAY;AACd,2BAAO,iBAAiB,MAAM;AAC9B,yBAAK,KAAK,iBAAiB;AAC3B,iCAAa;AAAA,kBACf;AAEA,yBAAO,iBAAiB,UAAU,KAAK;AAEvC,6BAAW,KAAK,QAAQ,MAAM,MAAM,KAAK,MAAM,GAAG;AAChD,2BAAO,kBAAmB,EAAE,oBAAoB,EAAE,aAAc;AAChE,0BAAM,KAAK,aAAa,aAAa,CAAC;AAAA,kBACxC;AAAA,gBACF;AAEA,oBAAI,CAAC,oBAAoB,CAAC,WAAW;AACnC,6BAAW,KAAK,QAAQ,MAAM,GAAG;AAC/B,2BAAO,kBAAmB,EAAE,oBAAoB,EAAE,aAAc;AAChE,0BAAM,KAAK,aAAa,aAAa,CAAC;AAAA,kBACxC;AAEA,yBAAO,iBAAiB,kBAAkB;AAE1C,wBAAM,KAAK,aAAa,eAAe;AAAA,gBACzC;AAEA,+BAAe;AAAA,cACjB,SAAS,OAAO;AACd,8BAAc,KAAK;AAAA,cACrB;AAAA,YACF,GAAG;AAAA,UACL,CAAC;AAEH,gBAAM,eAAe,SAAS;AAC9B,gBAAM,cAAc,QAAQ;AAE5B,cAAI;AACF,kBAAM,QAAQ,KAAK,CAAC,aAAa,OAAO,OAAO,KAAK,CAAC;AAAA,UACvD,UAAE;AACA,gBAAI,CAAC,YAAY,aAAa;AAC5B,oBAAM,iBAAiB,WAAW;AAAA,YACpC;AAEA,mBAAO,kBAAkB,OAAO,iBAAiB,KAAK,aAAa;AAEnE,gBAAI,OAAO,eAAe,YAAY,OAAO;AAC3C,mBAAK,aAAa,WAAW;AAAA,YAC/B;AAEA,gBAAI,CAAC,aAAa,aAAa;AAC7B,oBAAM,iBAAiB,YAAY;AAAA,YACrC;AAEA,gBAAI,CAAC,YAAY;AACf,kBAAI,CAAC,OAAO,aAAa;AACvB,uBAAO,iBAAiB,iBAAiB;AAAA,cAC3C;AAEA,mBAAK,KAAK,mBAAmB,OAAO,WAAW;AAAA,YACjD;AAEA,mBAAO,QAAQ,QAAQ;AACvB,kBAAM,OAAO,iBAAiB,MAAM,OAAO,WAAW;AAAA,UACxD;AAEA,kBAAQ;AAAA,QACV,SAAS,OAAO;AACd,iBAAO,KAAK;AAAA,QACd;AAAA,MACF,GAAG;AAAA,IACL,CAAC;AAAA,EACH;AACF;","names":[]}
1
+ {"version":3,"sources":["../../src/multimodal/agent_playout.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type AudioSource } from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport type { TextAudioSynchronizer } from '../transcription.js';\nimport { type AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';\n\nexport const proto = {};\n\nexport class PlayoutHandle extends EventEmitter {\n #audioSource: AudioSource;\n #sampleRate: number;\n #itemId: string;\n #contentIndex: number;\n /** @internal */\n synchronizer: TextAudioSynchronizer;\n /** @internal */\n doneFut: Future;\n /** @internal */\n intFut: Future;\n /** @internal */\n #interrupted: boolean;\n /** @internal */\n pushedDuration: number;\n /** @internal */\n totalPlayedTime: number | undefined; // Set when playout is done\n\n constructor(\n audioSource: AudioSource,\n sampleRate: number,\n itemId: string,\n contentIndex: number,\n synchronizer: TextAudioSynchronizer,\n ) {\n super();\n this.#audioSource = audioSource;\n this.#sampleRate = sampleRate;\n this.#itemId = itemId;\n this.#contentIndex = contentIndex;\n this.synchronizer = synchronizer;\n this.doneFut = new Future();\n this.intFut = new Future();\n this.#interrupted = false;\n this.pushedDuration = 0;\n this.totalPlayedTime = undefined;\n }\n\n get itemId(): string {\n return this.#itemId;\n }\n\n get audioSamples(): number {\n if (this.totalPlayedTime !== undefined) {\n return Math.floor(this.totalPlayedTime * this.#sampleRate);\n }\n\n return Math.floor(\n (this.pushedDuration - this.#audioSource.queuedDuration) * (this.#sampleRate / 1000),\n );\n }\n\n get textChars(): number {\n return this.synchronizer.playedText.length;\n }\n\n get contentIndex(): number {\n return this.#contentIndex;\n }\n\n get interrupted(): boolean {\n return this.#interrupted;\n }\n\n get done(): boolean {\n return this.doneFut.done || this.#interrupted;\n }\n\n interrupt() {\n if (this.doneFut.done) return;\n this.intFut.resolve();\n this.#interrupted = true;\n }\n}\n\nexport class AgentPlayout extends EventEmitter {\n #audioSource: AudioSource;\n #playoutTask: CancellablePromise<void> | null;\n #sampleRate: number;\n #numChannels: number;\n #inFrameSize: number;\n #outFrameSize: number;\n constructor(\n audioSource: AudioSource,\n sampleRate: number,\n numChannels: number,\n inFrameSize: number,\n outFrameSize: number,\n ) {\n super();\n this.#audioSource = audioSource;\n this.#playoutTask = null;\n this.#sampleRate = sampleRate;\n this.#numChannels = numChannels;\n this.#inFrameSize = inFrameSize;\n this.#outFrameSize = outFrameSize;\n }\n\n play(\n itemId: string,\n contentIndex: number,\n synchronizer: TextAudioSynchronizer,\n textStream: AsyncIterableQueue<string>,\n audioStream: AsyncIterableQueue<AudioFrame>,\n ): PlayoutHandle {\n const handle = new PlayoutHandle(\n this.#audioSource,\n this.#sampleRate,\n itemId,\n contentIndex,\n synchronizer,\n );\n this.#playoutTask = this.#makePlayoutTask(this.#playoutTask, handle, textStream, audioStream);\n return handle;\n }\n\n #makePlayoutTask(\n oldTask: CancellablePromise<void> | null,\n handle: PlayoutHandle,\n textStream: AsyncIterableQueue<string>,\n audioStream: AsyncIterableQueue<AudioFrame>,\n ): CancellablePromise<void> {\n return new CancellablePromise<void>((resolve, reject, onCancel) => {\n let cancelled = false;\n onCancel(() => {\n cancelled = true;\n });\n\n (async () => {\n try {\n if (oldTask) {\n await gracefullyCancel(oldTask);\n }\n\n let firstFrame = true;\n\n const readText = () =>\n new CancellablePromise<void>((resolveText, rejectText, onCancelText) => {\n let cancelledText = false;\n onCancelText(() => {\n cancelledText = true;\n });\n\n (async () => {\n try {\n for await (const text of textStream) {\n if (cancelledText || cancelled) {\n break;\n }\n handle.synchronizer.pushText(text);\n }\n handle.synchronizer.markTextSegmentEnd();\n resolveText();\n } catch (error) {\n rejectText(error);\n }\n })();\n });\n\n const capture = () =>\n new CancellablePromise<void>((resolveCapture, rejectCapture, onCancelCapture) => {\n let cancelledCapture = false;\n onCancelCapture(() => {\n cancelledCapture = true;\n });\n\n (async () => {\n try {\n const samplesPerChannel = this.#outFrameSize;\n const bstream = new AudioByteStream(\n this.#sampleRate,\n this.#numChannels,\n samplesPerChannel,\n );\n\n for await (const frame of audioStream) {\n if (cancelledCapture || cancelled) {\n break;\n }\n if (firstFrame) {\n handle.synchronizer.segmentPlayoutStarted();\n this.emit('playout_started');\n firstFrame = false;\n }\n\n handle.synchronizer.pushAudio(frame);\n\n for (const f of bstream.write(frame.data.buffer)) {\n handle.pushedDuration += (f.samplesPerChannel / f.sampleRate) * 1000;\n await this.#audioSource.captureFrame(f);\n }\n }\n\n if (!cancelledCapture && !cancelled) {\n for (const f of bstream.flush()) {\n handle.pushedDuration += (f.samplesPerChannel / f.sampleRate) * 1000;\n await this.#audioSource.captureFrame(f);\n }\n\n handle.synchronizer.markAudioSegmentEnd();\n\n await this.#audioSource.waitForPlayout();\n }\n\n resolveCapture();\n } catch (error) {\n rejectCapture(error);\n }\n })();\n });\n\n const readTextTask = readText();\n const captureTask = capture();\n\n try {\n await Promise.race([captureTask, handle.intFut.await]);\n } finally {\n if (!captureTask.isCancelled) {\n await gracefullyCancel(captureTask);\n }\n\n handle.totalPlayedTime = handle.pushedDuration - this.#audioSource.queuedDuration;\n\n if (handle.interrupted || captureTask.error) {\n await handle.synchronizer.close(true);\n this.#audioSource.clearQueue(); // make sure to remove any queued frames\n }\n\n if (!readTextTask.isCancelled) {\n await gracefullyCancel(readTextTask);\n }\n\n if (!firstFrame) {\n this.emit('playout_stopped', handle.interrupted);\n }\n\n handle.doneFut.resolve();\n await handle.synchronizer.close(false);\n }\n\n resolve();\n } catch (error) {\n reject(error);\n }\n })();\n });\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAC7B,SAAS,uBAAuB;AAEhC,SAAkC,oBAAoB,QAAQ,wBAAwB;AAE/E,MAAM,QAAQ,CAAC;AAEf,MAAM,sBAAsB,aAAa;AAAA,EAC9C;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA,YACE,aACA,YACA,QACA,cACA,cACA;AACA,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,UAAU;AACf,SAAK,gBAAgB;AACrB,SAAK,eAAe;AACpB,SAAK,UAAU,IAAI,OAAO;AAC1B,SAAK,SAAS,IAAI,OAAO;AACzB,SAAK,eAAe;AACpB,SAAK,iBAAiB;AACtB,SAAK,kBAAkB;AAAA,EACzB;AAAA,EAEA,IAAI,SAAiB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,QAAI,KAAK,oBAAoB,QAAW;AACtC,aAAO,KAAK,MAAM,KAAK,kBAAkB,KAAK,WAAW;AAAA,IAC3D;AAEA,WAAO,KAAK;AAAA,OACT,KAAK,iBAAiB,KAAK,aAAa,mBAAmB,KAAK,cAAc;AAAA,IACjF;AAAA,EACF;AAAA,EAEA,IAAI,YAAoB;AACtB,WAAO,KAAK,aAAa,WAAW;AAAA,EACtC;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAgB;AAClB,WAAO,KAAK,QAAQ,QAAQ,KAAK;AAAA,EACnC;AAAA,EAEA,YAAY;AACV,QAAI,KAAK,QAAQ,KAAM;AACvB,SAAK,OAAO,QAAQ;AACpB,SAAK,eAAe;AAAA,EACtB;AACF;AAEO,MAAM,qBAAqB,aAAa;AAAA,EAC7C;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,YACE,aACA,YACA,aACA,aACA,cACA;AACA,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,KACE,QACA,cACA,cACA,YACA,aACe;AACf,UAAM,SAAS,IAAI;AAAA,MACjB,KAAK;AAAA,MACL,KAAK;AAAA,MACL;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,SAAK,eAAe,KAAK,iBAAiB,KAAK,cAAc,QAAQ,YAAY,WAAW;AAC5F,WAAO;AAAA,EACT;AAAA,EAEA,iBACE,SACA,QACA,YACA,aAC0B;AAC1B,WAAO,IAAI,mBAAyB,CAAC,SAAS,QAAQ,aAAa;AACjE,UAAI,YAAY;AAChB,eAAS,MAAM;AACb,oBAAY;AAAA,MACd,CAAC;AAED,OAAC,YAAY;AACX,YAAI;AACF,cAAI,SAAS;AACX,kBAAM,iBAAiB,OAAO;AAAA,UAChC;AAEA,cAAI,aAAa;AAEjB,gBAAM,WAAW,MACf,IAAI,mBAAyB,CAAC,aAAa,YAAY,iBAAiB;AACtE,gBAAI,gBAAgB;AACpB,yBAAa,MAAM;AACjB,8BAAgB;AAAA,YAClB,CAAC;AAED,aAAC,YAAY;AACX,kBAAI;AACF,iCAAiB,QAAQ,YAAY;AACnC,sBAAI,iBAAiB,WAAW;AAC9B;AAAA,kBACF;AACA,yBAAO,aAAa,SAAS,IAAI;AAAA,gBACnC;AACA,uBAAO,aAAa,mBAAmB;AACvC,4BAAY;AAAA,cACd,SAAS,OAAO;AACd,2BAAW,KAAK;AAAA,cAClB;AAAA,YACF,GAAG;AAAA,UACL,CAAC;AAEH,gBAAM,UAAU,MACd,IAAI,mBAAyB,CAAC,gBAAgB,eAAe,oBAAoB;AAC/E,gBAAI,mBAAmB;AACvB,4BAAgB,MAAM;AACpB,iCAAmB;AAAA,YACrB,CAAC;AAED,aAAC,YAAY;AACX,kBAAI;AACF,sBAAM,oBAAoB,KAAK;AAC/B,sBAAM,UAAU,IAAI;AAAA,kBAClB,KAAK;AAAA,kBACL,KAAK;AAAA,kBACL;AAAA,gBACF;AAEA,iCAAiB,SAAS,aAAa;AACrC,sBAAI,oBAAoB,WAAW;AACjC;AAAA,kBACF;AACA,sBAAI,YAAY;AACd,2BAAO,aAAa,sBAAsB;AAC1C,yBAAK,KAAK,iBAAiB;AAC3B,iCAAa;AAAA,kBACf;AAEA,yBAAO,aAAa,UAAU,KAAK;AAEnC,6BAAW,KAAK,QAAQ,MAAM,MAAM,KAAK,MAAM,GAAG;AAChD,2BAAO,kBAAmB,EAAE,oBAAoB,EAAE,aAAc;AAChE,0BAAM,KAAK,aAAa,aAAa,CAAC;AAAA,kBACxC;AAAA,gBACF;AAEA,oBAAI,CAAC,oBAAoB,CAAC,WAAW;AACnC,6BAAW,KAAK,QAAQ,MAAM,GAAG;AAC/B,2BAAO,kBAAmB,EAAE,oBAAoB,EAAE,aAAc;AAChE,0BAAM,KAAK,aAAa,aAAa,CAAC;AAAA,kBACxC;AAEA,yBAAO,aAAa,oBAAoB;AAExC,wBAAM,KAAK,aAAa,eAAe;AAAA,gBACzC;AAEA,+BAAe;AAAA,cACjB,SAAS,OAAO;AACd,8BAAc,KAAK;AAAA,cACrB;AAAA,YACF,GAAG;AAAA,UACL,CAAC;AAEH,gBAAM,eAAe,SAAS;AAC9B,gBAAM,cAAc,QAAQ;AAE5B,cAAI;AACF,kBAAM,QAAQ,KAAK,CAAC,aAAa,OAAO,OAAO,KAAK,CAAC;AAAA,UACvD,UAAE;AACA,gBAAI,CAAC,YAAY,aAAa;AAC5B,oBAAM,iBAAiB,WAAW;AAAA,YACpC;AAEA,mBAAO,kBAAkB,OAAO,iBAAiB,KAAK,aAAa;AAEnE,gBAAI,OAAO,eAAe,YAAY,OAAO;AAC3C,oBAAM,OAAO,aAAa,MAAM,IAAI;AACpC,mBAAK,aAAa,WAAW;AAAA,YAC/B;AAEA,gBAAI,CAAC,aAAa,aAAa;AAC7B,oBAAM,iBAAiB,YAAY;AAAA,YACrC;AAEA,gBAAI,CAAC,YAAY;AACf,mBAAK,KAAK,mBAAmB,OAAO,WAAW;AAAA,YACjD;AAEA,mBAAO,QAAQ,QAAQ;AACvB,kBAAM,OAAO,aAAa,MAAM,KAAK;AAAA,UACvC;AAEA,kBAAQ;AAAA,QACV,SAAS,OAAO;AACd,iBAAO,KAAK;AAAA,QACd;AAAA,MACF,GAAG;AAAA,IACL,CAAC;AAAA,EACH;AACF;","names":[]}
@@ -150,7 +150,7 @@ class MultimodalAgent extends import_node_events.EventEmitter {
150
150
  this.emit("agent_stopped_speaking");
151
151
  this.#speaking = false;
152
152
  if (this.#playingHandle) {
153
- let text = this.#playingHandle.transcriptionFwd.text;
153
+ let text = this.#playingHandle.synchronizer.playedText;
154
154
  if (interrupted) {
155
155
  text += "\u2026";
156
156
  }
@@ -195,16 +195,20 @@ class MultimodalAgent extends import_node_events.EventEmitter {
195
195
  this.#session.on("response_content_added", (message) => {
196
196
  var _a2;
197
197
  if (message.contentType === "text") return;
198
- const trFwd = new import_transcription.BasicTranscriptionForwarder(
199
- this.room,
200
- this.room.localParticipant.identity,
201
- this.#getLocalTrackSid(),
202
- message.responseId
203
- );
198
+ const synchronizer = new import_transcription.TextAudioSynchronizer(import_transcription.defaultTextSyncOptions);
199
+ synchronizer.on("textUpdated", (text) => {
200
+ this.#publishTranscription(
201
+ this.room.localParticipant.identity,
202
+ this.#getLocalTrackSid(),
203
+ text.text,
204
+ text.final,
205
+ text.id
206
+ );
207
+ });
204
208
  const handle = (_a2 = this.#agentPlayout) == null ? void 0 : _a2.play(
205
209
  message.itemId,
206
210
  message.contentIndex,
207
- trFwd,
211
+ synchronizer,
208
212
  message.textStream,
209
213
  message.audioStream
210
214
  );
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/multimodal/multimodal_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrack,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n AudioStream,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport * as llm from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { MultimodalLLMMetrics } from '../metrics/base.js';\nimport { BasicTranscriptionForwarder } from '../transcription.js';\nimport { findMicroTrackId } from '../utils.js';\nimport { AgentPlayout, type PlayoutHandle } from './agent_playout.js';\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeSession extends EventEmitter {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract conversation: any; // openai.realtime.Conversation\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract inputAudioBuffer: any; // openai.realtime.InputAudioBuffer\n abstract fncCtx: llm.FunctionContext | undefined;\n abstract recoverFromTextResponse(itemId: string): void;\n}\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeModel {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract session(options: any): RealtimeSession; // openai.realtime.ModelOptions\n abstract close(): Promise<void>;\n abstract sampleRate: number;\n abstract numChannels: number;\n abstract inFrameSize: number;\n abstract outFrameSize: number;\n}\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\n\n/** @beta */\nexport class MultimodalAgent extends EventEmitter {\n model: RealtimeModel;\n room: Room | null = null;\n linkedParticipant: RemoteParticipant | null = null;\n subscribedTrack: RemoteAudioTrack | null = null;\n readMicroTask: Promise<void> | null = null;\n\n #textResponseRetries = 0;\n #maxTextResponseRetries: number;\n\n constructor({\n model,\n chatCtx,\n fncCtx,\n maxTextResponseRetries = 5,\n }: {\n model: RealtimeModel;\n chatCtx?: llm.ChatContext;\n fncCtx?: llm.FunctionContext;\n maxTextResponseRetries?: number;\n }) {\n super();\n this.model = model;\n this.#chatCtx = chatCtx;\n this.#fncCtx = fncCtx;\n this.#maxTextResponseRetries = maxTextResponseRetries;\n }\n\n #participant: RemoteParticipant | string | null = null;\n #agentPublication: LocalTrackPublication | null = null;\n #localTrackSid: string | null = null;\n #localSource: AudioSource | null = null;\n #agentPlayout: AgentPlayout | null = null;\n #playingHandle: PlayoutHandle | undefined = undefined;\n #logger = log();\n #session: RealtimeSession | null = null;\n #fncCtx: llm.FunctionContext | undefined = undefined;\n #chatCtx: llm.ChatContext | undefined = undefined;\n\n #_started: boolean = false;\n #_pendingFunctionCalls: Set<string> = new Set();\n #_speaking: boolean = false;\n\n get fncCtx(): llm.FunctionContext | undefined {\n return this.#fncCtx;\n }\n\n set fncCtx(ctx: llm.FunctionContext | undefined) {\n this.#fncCtx = ctx;\n if (this.#session) {\n this.#session.fncCtx = ctx;\n }\n }\n\n get #pendingFunctionCalls(): Set<string> {\n return this.#_pendingFunctionCalls;\n }\n\n set #pendingFunctionCalls(calls: Set<string>) {\n this.#_pendingFunctionCalls = calls;\n this.#updateState();\n }\n\n get #speaking(): boolean {\n return this.#_speaking;\n }\n\n set #speaking(isSpeaking: boolean) {\n this.#_speaking = isSpeaking;\n this.#updateState();\n }\n\n get #started(): boolean {\n return this.#_started;\n }\n\n set #started(started: boolean) {\n this.#_started = started;\n this.#updateState();\n }\n\n start(\n room: Room,\n participant: RemoteParticipant | string | null = null,\n ): Promise<RealtimeSession> {\n return new Promise(async (resolve, reject) => {\n if (this.#started) {\n reject(new Error('MultimodalAgent already started'));\n }\n this.#updateState();\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.linkedParticipant) {\n return;\n }\n this.#linkParticipant(participant.identity!);\n });\n room.on(\n RoomEvent.TrackPublished,\n (trackPublication: RemoteTrackPublication, participant: RemoteParticipant) => {\n if (\n this.linkedParticipant &&\n participant.identity === this.linkedParticipant.identity &&\n trackPublication.source === TrackSource.SOURCE_MICROPHONE &&\n !trackPublication.subscribed\n ) {\n trackPublication.setSubscribed(true);\n }\n },\n );\n room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));\n\n this.room = room;\n this.#participant = participant;\n\n this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);\n this.#agentPlayout = new AgentPlayout(\n this.#localSource,\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n this.model.outFrameSize,\n );\n const onPlayoutStarted = () => {\n this.emit('agent_started_speaking');\n this.#speaking = true;\n };\n\n const onPlayoutStopped = (interrupted: boolean) => {\n this.emit('agent_stopped_speaking');\n this.#speaking = false;\n if (this.#playingHandle) {\n let text = this.#playingHandle.transcriptionFwd.text;\n if (interrupted) {\n text += '…';\n }\n const msg = llm.ChatMessage.create({\n role: llm.ChatRole.ASSISTANT,\n text,\n });\n\n if (interrupted) {\n this.emit('agent_speech_interrupted', msg);\n } else {\n this.emit('agent_speech_committed', msg);\n }\n this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');\n }\n };\n\n this.#agentPlayout.on('playout_started', onPlayoutStarted);\n this.#agentPlayout.on('playout_stopped', onPlayoutStopped);\n\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);\n const options = new TrackPublishOptions();\n options.source = TrackSource.SOURCE_MICROPHONE;\n this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;\n if (!this.#agentPublication) {\n this.#logger.error('Failed to publish track');\n reject(new Error('Failed to publish track'));\n return;\n }\n\n await this.#agentPublication.waitForSubscription();\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity!);\n }\n } else {\n // No participant specified, try to find the first participant in the room\n for (const participant of room.remoteParticipants.values()) {\n this.#linkParticipant(participant.identity!);\n break;\n }\n }\n\n this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });\n this.#started = true;\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_added', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') return;\n\n const trFwd = new BasicTranscriptionForwarder(\n this.room!,\n this.room!.localParticipant!.identity!,\n this.#getLocalTrackSid()!,\n message.responseId,\n );\n\n const handle = this.#agentPlayout?.play(\n message.itemId,\n message.contentIndex,\n trFwd,\n message.textStream,\n message.audioStream,\n );\n this.#playingHandle = handle;\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_done', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') {\n if (this.#textResponseRetries >= this.#maxTextResponseRetries) {\n throw new Error(\n 'The OpenAI Realtime API returned a text response ' +\n `after ${this.#maxTextResponseRetries} retries. ` +\n 'Please try to reduce the number of text system or ' +\n 'assistant messages in the chat context.',\n );\n }\n\n this.#textResponseRetries++;\n this.#logger\n .child({\n itemId: message.itemId,\n text: message.text,\n retries: this.#textResponseRetries,\n })\n .warn(\n 'The OpenAI Realtime API returned a text response instead of audio. ' +\n 'Attempting to recover to audio mode...',\n );\n this.#session!.recoverFromTextResponse(message.itemId);\n } else {\n this.#textResponseRetries = 0;\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_committed', (ev: any) => {\n // openai.realtime.InputSpeechCommittedEvent\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_transcription_completed', (ev: any) => {\n // openai.realtime.InputSpeechTranscriptionCompletedEvent\n const transcription = ev.transcript;\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, transcription, true, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n const userMsg = llm.ChatMessage.create({\n role: llm.ChatRole.USER,\n text: transcription,\n });\n this.emit('user_speech_committed', userMsg);\n this.#logger.child({ transcription }).debug('committed user speech');\n });\n\n this.#session.on('input_speech_started', (ev: any) => {\n this.emit('user_started_speaking');\n if (this.#playingHandle && !this.#playingHandle.done) {\n this.#playingHandle.interrupt();\n\n this.#session!.conversation.item.truncate(\n this.#playingHandle.itemId,\n this.#playingHandle.contentIndex,\n Math.floor((this.#playingHandle.audioSamples / 24000) * 1000),\n );\n\n this.#playingHandle = undefined;\n }\n\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#session.on('input_speech_stopped', (ev: any) => {\n this.emit('user_stopped_speaking');\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_started', (ev: any) => {\n this.#pendingFunctionCalls.add(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_completed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_failed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n this.#session.on('metrics_collected', (metrics: MultimodalLLMMetrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n resolve(this.#session);\n });\n }\n\n #linkParticipant(participantIdentity: string): void {\n if (!this.room) {\n this.#logger.error('Room is not set');\n return;\n }\n\n this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;\n if (!this.linkedParticipant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n if (this.linkedParticipant.trackPublications.size > 0) {\n this.#subscribeToMicrophone();\n }\n\n // also check if already subscribed\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {\n this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);\n break;\n }\n }\n }\n\n #subscribeToMicrophone(): void {\n if (!this.linkedParticipant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n }\n\n #handleTrackSubscription(\n track: RemoteTrack,\n publication: RemoteTrackPublication,\n participant: RemoteParticipant,\n ) {\n if (\n publication.source !== TrackSource.SOURCE_MICROPHONE ||\n participant.identity !== this.linkedParticipant?.identity\n ) {\n return;\n }\n const readAudioStreamTask = async (audioStream: AudioStream) => {\n const bstream = new AudioByteStream(\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n );\n\n for await (const frame of audioStream) {\n const audioData = frame.data;\n for (const frame of bstream.write(audioData.buffer)) {\n this.#session!.inputAudioBuffer.append(frame);\n }\n }\n };\n this.subscribedTrack = track;\n\n this.readMicroTask = new Promise<void>((resolve, reject) => {\n readAudioStreamTask(new AudioStream(track, this.model.sampleRate, this.model.numChannels))\n .then(resolve)\n .catch(reject);\n });\n }\n\n #getLocalTrackSid(): string | null {\n if (!this.#localTrackSid && this.room && this.room.localParticipant) {\n this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant!.identity!);\n }\n return this.#localTrackSid;\n }\n\n #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ): void {\n this.#logger.debug(\n `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,\n );\n if (!this.room?.localParticipant) {\n this.#logger.error('Room or local participant not set');\n return;\n }\n\n this.room.localParticipant.publishTranscription({\n participantIdentity,\n trackSid,\n segments: [\n {\n text,\n final: isFinal,\n id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n }\n\n #updateState() {\n let newState: AgentState = 'initializing';\n if (this.#pendingFunctionCalls.size > 0) {\n newState = 'thinking';\n } else if (this.#speaking) {\n newState = 'speaking';\n } else if (this.#started) {\n newState = 'listening';\n }\n\n this.#setState(newState);\n }\n\n #setState(state: AgentState) {\n if (this.room?.isConnected && this.room.localParticipant) {\n const currentState = this.room.localParticipant.attributes![AGENT_STATE_ATTRIBUTE];\n if (currentState !== state) {\n this.room.localParticipant.setAttributes({\n [AGENT_STATE_ATTRIBUTE]: state,\n });\n this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWA,sBAOO;AACP,yBAA6B;AAC7B,mBAAgC;AAChC,UAAqB;AACrB,iBAAoB;AAEpB,2BAA4C;AAC5C,mBAAiC;AACjC,2BAAiD;AAM1C,MAAe,wBAAwB,gCAAa;AAO3D;AAMO,MAAe,cAAc;AAQpC;AAGO,MAAM,wBAAwB;AAG9B,MAAM,wBAAwB,gCAAa;AAAA,EAChD;AAAA,EACA,OAAoB;AAAA,EACpB,oBAA8C;AAAA,EAC9C,kBAA2C;AAAA,EAC3C,gBAAsC;AAAA,EAEtC,uBAAuB;AAAA,EACvB;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA,yBAAyB;AAAA,EAC3B,GAKG;AACD,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAU;AACf,SAAK,0BAA0B;AAAA,EACjC;AAAA,EAEA,eAAkD;AAAA,EAClD,oBAAkD;AAAA,EAClD,iBAAgC;AAAA,EAChC,eAAmC;AAAA,EACnC,gBAAqC;AAAA,EACrC,iBAA4C;AAAA,EAC5C,cAAU,gBAAI;AAAA,EACd,WAAmC;AAAA,EACnC,UAA2C;AAAA,EAC3C,WAAwC;AAAA,EAExC,YAAqB;AAAA,EACrB,yBAAsC,oBAAI,IAAI;AAAA,EAC9C,aAAsB;AAAA,EAEtB,IAAI,SAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAO,KAAsC;AAC/C,SAAK,UAAU;AACf,QAAI,KAAK,UAAU;AACjB,WAAK,SAAS,SAAS;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,IAAI,wBAAqC;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAAsB,OAAoB;AAC5C,SAAK,yBAAyB;AAC9B,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAU,YAAqB;AACjC,SAAK,aAAa;AAClB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,SAAkB;AAC7B,SAAK,YAAY;AACjB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,MACE,MACA,cAAiD,MACvB;AAC1B,WAAO,IAAI,QAAQ,OAAO,SAAS,WAAW;AAhJlD;AAiJM,UAAI,KAAK,UAAU;AACjB,eAAO,IAAI,MAAM,iCAAiC,CAAC;AAAA,MACrD;AACA,WAAK,aAAa;AAElB,WAAK,GAAG,0BAAU,sBAAsB,CAACA,iBAAmC;AAE1E,YAAI,KAAK,mBAAmB;AAC1B;AAAA,QACF;AACA,aAAK,iBAAiBA,aAAY,QAAS;AAAA,MAC7C,CAAC;AACD,WAAK;AAAA,QACH,0BAAU;AAAA,QACV,CAAC,kBAA0CA,iBAAmC;AAC5E,cACE,KAAK,qBACLA,aAAY,aAAa,KAAK,kBAAkB,YAChD,iBAAiB,WAAW,4BAAY,qBACxC,CAAC,iBAAiB,YAClB;AACA,6BAAiB,cAAc,IAAI;AAAA,UACrC;AAAA,QACF;AAAA,MACF;AACA,WAAK,GAAG,0BAAU,iBAAiB,KAAK,yBAAyB,KAAK,IAAI,CAAC;AAE3E,WAAK,OAAO;AACZ,WAAK,eAAe;AAEpB,WAAK,eAAe,IAAI,4BAAY,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW;AACjF,WAAK,gBAAgB,IAAI;AAAA,QACvB,KAAK;AAAA,QACL,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AACA,YAAM,mBAAmB,MAAM;AAC7B,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AAAA,MACnB;AAEA,YAAM,mBAAmB,CAAC,gBAAyB;AACjD,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AACjB,YAAI,KAAK,gBAAgB;AACvB,cAAI,OAAO,KAAK,eAAe,iBAAiB;AAChD,cAAI,aAAa;AACf,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,YAAY,OAAO;AAAA,YACjC,MAAM,IAAI,SAAS;AAAA,YACnB;AAAA,UACF,CAAC;AAED,cAAI,aAAa;AACf,iBAAK,KAAK,4BAA4B,GAAG;AAAA,UAC3C,OAAO;AACL,iBAAK,KAAK,0BAA0B,GAAG;AAAA,UACzC;AACA,eAAK,QAAQ,MAAM,EAAE,eAAe,MAAM,YAAY,CAAC,EAAE,MAAM,wBAAwB;AAAA,QACzF;AAAA,MACF;AAEA,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AACzD,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AAEzD,YAAM,QAAQ,gCAAgB,iBAAiB,mBAAmB,KAAK,YAAY;AACnF,YAAM,UAAU,IAAI,oCAAoB;AACxC,cAAQ,SAAS,4BAAY;AAC7B,WAAK,oBAAqB,QAAM,UAAK,qBAAL,mBAAuB,aAAa,OAAO,aAAa;AACxF,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,QAAQ,MAAM,yBAAyB;AAC5C,eAAO,IAAI,MAAM,yBAAyB,CAAC;AAC3C;AAAA,MACF;AAEA,YAAM,KAAK,kBAAkB,oBAAoB;AAEjD,UAAI,aAAa;AACf,YAAI,OAAO,gBAAgB,UAAU;AACnC,eAAK,iBAAiB,WAAW;AAAA,QACnC,OAAO;AACL,eAAK,iBAAiB,YAAY,QAAS;AAAA,QAC7C;AAAA,MACF,OAAO;AAEL,mBAAWA,gBAAe,KAAK,mBAAmB,OAAO,GAAG;AAC1D,eAAK,iBAAiBA,aAAY,QAAS;AAC3C;AAAA,QACF;AAAA,MACF;AAEA,WAAK,WAAW,KAAK,MAAM,QAAQ,EAAE,QAAQ,KAAK,SAAS,SAAS,KAAK,SAAS,CAAC;AACnF,WAAK,WAAW;AAGhB,WAAK,SAAS,GAAG,0BAA0B,CAAC,YAAiB;AAnPnE,YAAAC;AAqPQ,YAAI,QAAQ,gBAAgB,OAAQ;AAEpC,cAAM,QAAQ,IAAI;AAAA,UAChB,KAAK;AAAA,UACL,KAAK,KAAM,iBAAkB;AAAA,UAC7B,KAAK,kBAAkB;AAAA,UACvB,QAAQ;AAAA,QACV;AAEA,cAAM,UAASA,MAAA,KAAK,kBAAL,gBAAAA,IAAoB;AAAA,UACjC,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA,QAAQ;AAAA,UACR,QAAQ;AAAA;AAEV,aAAK,iBAAiB;AAAA,MACxB,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,YAAiB;AAE1D,YAAI,QAAQ,gBAAgB,QAAQ;AAClC,cAAI,KAAK,wBAAwB,KAAK,yBAAyB;AAC7D,kBAAM,IAAI;AAAA,cACR,0DACW,KAAK,uBAAuB;AAAA,YAGzC;AAAA,UACF;AAEA,eAAK;AACL,eAAK,QACF,MAAM;AAAA,YACL,QAAQ,QAAQ;AAAA,YAChB,MAAM,QAAQ;AAAA,YACd,SAAS,KAAK;AAAA,UAChB,CAAC,EACA;AAAA,YACC;AAAA,UAEF;AACF,eAAK,SAAU,wBAAwB,QAAQ,MAAM;AAAA,QACvD,OAAO;AACL,eAAK,uBAAuB;AAAA,QAC9B;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,0BAA0B,CAAC,OAAY;AAvS9D,YAAAA,KAAA;AAySQ,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wCAAwC,CAAC,OAAY;AAnT5E,YAAAA,KAAA;AAqTQ,cAAM,gBAAgB,GAAG;AACzB,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,eAAe,MAAM,GAAG,MAAM;AAAA,QAC1F,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AACA,cAAM,UAAU,IAAI,YAAY,OAAO;AAAA,UACrC,MAAM,IAAI,SAAS;AAAA,UACnB,MAAM;AAAA,QACR,CAAC;AACD,aAAK,KAAK,yBAAyB,OAAO;AAC1C,aAAK,QAAQ,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,uBAAuB;AAAA,MACrE,CAAC;AAED,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AArU5D,YAAAA,KAAA;AAsUQ,aAAK,KAAK,uBAAuB;AACjC,YAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,MAAM;AACpD,eAAK,eAAe,UAAU;AAE9B,eAAK,SAAU,aAAa,KAAK;AAAA,YAC/B,KAAK,eAAe;AAAA,YACpB,KAAK,eAAe;AAAA,YACpB,KAAK,MAAO,KAAK,eAAe,eAAe,OAAS,GAAI;AAAA,UAC9D;AAEA,eAAK,iBAAiB;AAAA,QACxB;AAEA,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,KAAK,uBAAuB;AAAA,MACnC,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,OAAY;AACrD,aAAK,sBAAsB,IAAI,GAAG,MAAM;AACxC,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,2BAA2B,CAAC,OAAY;AACvD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAED,WAAK,SAAS,GAAG,qBAAqB,CAAC,YAAkC;AACvE,aAAK,KAAK,qBAAqB,OAAO;AAAA,MACxC,CAAC;AAED,cAAQ,KAAK,QAAQ;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,MAAM;AACd,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,oBAAoB,KAAK,KAAK,mBAAmB,IAAI,mBAAmB,KAAK;AAClF,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,QAAI,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AACrD,WAAK,uBAAuB;AAAA,IAC9B;AAGA,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,4BAAY,qBAAqB,YAAY,OAAO;AAC7E,aAAK,yBAAyB,YAAY,OAAO,aAAa,KAAK,iBAAiB;AACpF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,4BAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAAA,EACF;AAAA,EAEA,yBACE,OACA,aACA,aACA;AA5aJ;AA6aI,QACE,YAAY,WAAW,4BAAY,qBACnC,YAAY,eAAa,UAAK,sBAAL,mBAAwB,WACjD;AACA;AAAA,IACF;AACA,UAAM,sBAAsB,OAAO,gBAA6B;AAC9D,YAAM,UAAU,IAAI;AAAA,QAClB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AAEA,uBAAiB,SAAS,aAAa;AACrC,cAAM,YAAY,MAAM;AACxB,mBAAWC,UAAS,QAAQ,MAAM,UAAU,MAAM,GAAG;AACnD,eAAK,SAAU,iBAAiB,OAAOA,MAAK;AAAA,QAC9C;AAAA,MACF;AAAA,IACF;AACA,SAAK,kBAAkB;AAEvB,SAAK,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC1D,0BAAoB,IAAI,4BAAY,OAAO,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW,CAAC,EACtF,KAAK,OAAO,EACZ,MAAM,MAAM;AAAA,IACjB,CAAC;AAAA,EACH;AAAA,EAEA,oBAAmC;AACjC,QAAI,CAAC,KAAK,kBAAkB,KAAK,QAAQ,KAAK,KAAK,kBAAkB;AACnE,WAAK,qBAAiB,+BAAiB,KAAK,MAAM,KAAK,KAAK,iBAAkB,QAAS;AAAA,IACzF;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,sBACE,qBACA,UACA,MACA,SACA,IACM;AAvdV;AAwdI,SAAK,QAAQ;AAAA,MACX,4BAA4B,mBAAmB,IAAI,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,EAAE;AAAA,IACtF;AACA,QAAI,GAAC,UAAK,SAAL,mBAAW,mBAAkB;AAChC,WAAK,QAAQ,MAAM,mCAAmC;AACtD;AAAA,IACF;AAEA,SAAK,KAAK,iBAAiB,qBAAqB;AAAA,MAC9C;AAAA,MACA;AAAA,MACA,UAAU;AAAA,QACR;AAAA,UACE;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,eAAe;AACb,QAAI,WAAuB;AAC3B,QAAI,KAAK,sBAAsB,OAAO,GAAG;AACvC,iBAAW;AAAA,IACb,WAAW,KAAK,WAAW;AACzB,iBAAW;AAAA,IACb,WAAW,KAAK,UAAU;AACxB,iBAAW;AAAA,IACb;AAEA,SAAK,UAAU,QAAQ;AAAA,EACzB;AAAA,EAEA,UAAU,OAAmB;AA7f/B;AA8fI,UAAI,UAAK,SAAL,mBAAW,gBAAe,KAAK,KAAK,kBAAkB;AACxD,YAAM,eAAe,KAAK,KAAK,iBAAiB,WAAY,qBAAqB;AACjF,UAAI,iBAAiB,OAAO;AAC1B,aAAK,KAAK,iBAAiB,cAAc;AAAA,UACvC,CAAC,qBAAqB,GAAG;AAAA,QAC3B,CAAC;AACD,aAAK,QAAQ,MAAM,GAAG,qBAAqB,KAAK,YAAY,MAAM,KAAK,EAAE;AAAA,MAC3E;AAAA,IACF;AAAA,EACF;AACF;","names":["participant","_a","frame"]}
1
+ {"version":3,"sources":["../../src/multimodal/multimodal_agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n LocalTrackPublication,\n RemoteAudioTrack,\n RemoteParticipant,\n RemoteTrack,\n RemoteTrackPublication,\n Room,\n} from '@livekit/rtc-node';\nimport {\n AudioSource,\n AudioStream,\n LocalAudioTrack,\n RoomEvent,\n TrackPublishOptions,\n TrackSource,\n} from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport { AudioByteStream } from '../audio.js';\nimport * as llm from '../llm/index.js';\nimport { log } from '../log.js';\nimport type { MultimodalLLMMetrics } from '../metrics/base.js';\nimport { TextAudioSynchronizer, defaultTextSyncOptions } from '../transcription.js';\nimport { findMicroTrackId } from '../utils.js';\nimport { AgentPlayout, type PlayoutHandle } from './agent_playout.js';\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeSession extends EventEmitter {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract conversation: any; // openai.realtime.Conversation\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract inputAudioBuffer: any; // openai.realtime.InputAudioBuffer\n abstract fncCtx: llm.FunctionContext | undefined;\n abstract recoverFromTextResponse(itemId: string): void;\n}\n\n/**\n * @internal\n * @beta\n */\nexport abstract class RealtimeModel {\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n abstract session(options: any): RealtimeSession; // openai.realtime.ModelOptions\n abstract close(): Promise<void>;\n abstract sampleRate: number;\n abstract numChannels: number;\n abstract inFrameSize: number;\n abstract outFrameSize: number;\n}\n\nexport type AgentState = 'initializing' | 'thinking' | 'listening' | 'speaking';\nexport const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';\n\n/** @beta */\nexport class MultimodalAgent extends EventEmitter {\n model: RealtimeModel;\n room: Room | null = null;\n linkedParticipant: RemoteParticipant | null = null;\n subscribedTrack: RemoteAudioTrack | null = null;\n readMicroTask: Promise<void> | null = null;\n\n #textResponseRetries = 0;\n #maxTextResponseRetries: number;\n\n constructor({\n model,\n chatCtx,\n fncCtx,\n maxTextResponseRetries = 5,\n }: {\n model: RealtimeModel;\n chatCtx?: llm.ChatContext;\n fncCtx?: llm.FunctionContext;\n maxTextResponseRetries?: number;\n }) {\n super();\n this.model = model;\n this.#chatCtx = chatCtx;\n this.#fncCtx = fncCtx;\n this.#maxTextResponseRetries = maxTextResponseRetries;\n }\n\n #participant: RemoteParticipant | string | null = null;\n #agentPublication: LocalTrackPublication | null = null;\n #localTrackSid: string | null = null;\n #localSource: AudioSource | null = null;\n #agentPlayout: AgentPlayout | null = null;\n #playingHandle: PlayoutHandle | undefined = undefined;\n #logger = log();\n #session: RealtimeSession | null = null;\n #fncCtx: llm.FunctionContext | undefined = undefined;\n #chatCtx: llm.ChatContext | undefined = undefined;\n\n #_started: boolean = false;\n #_pendingFunctionCalls: Set<string> = new Set();\n #_speaking: boolean = false;\n\n get fncCtx(): llm.FunctionContext | undefined {\n return this.#fncCtx;\n }\n\n set fncCtx(ctx: llm.FunctionContext | undefined) {\n this.#fncCtx = ctx;\n if (this.#session) {\n this.#session.fncCtx = ctx;\n }\n }\n\n get #pendingFunctionCalls(): Set<string> {\n return this.#_pendingFunctionCalls;\n }\n\n set #pendingFunctionCalls(calls: Set<string>) {\n this.#_pendingFunctionCalls = calls;\n this.#updateState();\n }\n\n get #speaking(): boolean {\n return this.#_speaking;\n }\n\n set #speaking(isSpeaking: boolean) {\n this.#_speaking = isSpeaking;\n this.#updateState();\n }\n\n get #started(): boolean {\n return this.#_started;\n }\n\n set #started(started: boolean) {\n this.#_started = started;\n this.#updateState();\n }\n\n start(\n room: Room,\n participant: RemoteParticipant | string | null = null,\n ): Promise<RealtimeSession> {\n return new Promise(async (resolve, reject) => {\n if (this.#started) {\n reject(new Error('MultimodalAgent already started'));\n }\n this.#updateState();\n\n room.on(RoomEvent.ParticipantConnected, (participant: RemoteParticipant) => {\n // automatically link to the first participant that connects, if not already linked\n if (this.linkedParticipant) {\n return;\n }\n this.#linkParticipant(participant.identity!);\n });\n room.on(\n RoomEvent.TrackPublished,\n (trackPublication: RemoteTrackPublication, participant: RemoteParticipant) => {\n if (\n this.linkedParticipant &&\n participant.identity === this.linkedParticipant.identity &&\n trackPublication.source === TrackSource.SOURCE_MICROPHONE &&\n !trackPublication.subscribed\n ) {\n trackPublication.setSubscribed(true);\n }\n },\n );\n room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));\n\n this.room = room;\n this.#participant = participant;\n\n this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);\n this.#agentPlayout = new AgentPlayout(\n this.#localSource,\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n this.model.outFrameSize,\n );\n const onPlayoutStarted = () => {\n this.emit('agent_started_speaking');\n this.#speaking = true;\n };\n\n const onPlayoutStopped = (interrupted: boolean) => {\n this.emit('agent_stopped_speaking');\n this.#speaking = false;\n if (this.#playingHandle) {\n let text = this.#playingHandle.synchronizer.playedText;\n if (interrupted) {\n text += '…';\n }\n const msg = llm.ChatMessage.create({\n role: llm.ChatRole.ASSISTANT,\n text,\n });\n\n if (interrupted) {\n this.emit('agent_speech_interrupted', msg);\n } else {\n this.emit('agent_speech_committed', msg);\n }\n this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');\n }\n };\n\n this.#agentPlayout.on('playout_started', onPlayoutStarted);\n this.#agentPlayout.on('playout_stopped', onPlayoutStopped);\n\n const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);\n const options = new TrackPublishOptions();\n options.source = TrackSource.SOURCE_MICROPHONE;\n this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;\n if (!this.#agentPublication) {\n this.#logger.error('Failed to publish track');\n reject(new Error('Failed to publish track'));\n return;\n }\n\n await this.#agentPublication.waitForSubscription();\n\n if (participant) {\n if (typeof participant === 'string') {\n this.#linkParticipant(participant);\n } else {\n this.#linkParticipant(participant.identity!);\n }\n } else {\n // No participant specified, try to find the first participant in the room\n for (const participant of room.remoteParticipants.values()) {\n this.#linkParticipant(participant.identity!);\n break;\n }\n }\n\n this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });\n this.#started = true;\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_added', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') return;\n\n const synchronizer = new TextAudioSynchronizer(defaultTextSyncOptions);\n synchronizer.on('textUpdated', (text) => {\n this.#publishTranscription(\n this.room!.localParticipant!.identity!,\n this.#getLocalTrackSid()!,\n text.text,\n text.final,\n text.id,\n );\n });\n\n const handle = this.#agentPlayout?.play(\n message.itemId,\n message.contentIndex,\n synchronizer,\n message.textStream,\n message.audioStream,\n );\n this.#playingHandle = handle;\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('response_content_done', (message: any) => {\n // openai.realtime.RealtimeContent\n if (message.contentType === 'text') {\n if (this.#textResponseRetries >= this.#maxTextResponseRetries) {\n throw new Error(\n 'The OpenAI Realtime API returned a text response ' +\n `after ${this.#maxTextResponseRetries} retries. ` +\n 'Please try to reduce the number of text system or ' +\n 'assistant messages in the chat context.',\n );\n }\n\n this.#textResponseRetries++;\n this.#logger\n .child({\n itemId: message.itemId,\n text: message.text,\n retries: this.#textResponseRetries,\n })\n .warn(\n 'The OpenAI Realtime API returned a text response instead of audio. ' +\n 'Attempting to recover to audio mode...',\n );\n this.#session!.recoverFromTextResponse(message.itemId);\n } else {\n this.#textResponseRetries = 0;\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_committed', (ev: any) => {\n // openai.realtime.InputSpeechCommittedEvent\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('input_speech_transcription_completed', (ev: any) => {\n // openai.realtime.InputSpeechTranscriptionCompletedEvent\n const transcription = ev.transcript;\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, transcription, true, ev.itemId);\n } else {\n this.#logger.error('Participant or track not set');\n }\n const userMsg = llm.ChatMessage.create({\n role: llm.ChatRole.USER,\n text: transcription,\n });\n this.emit('user_speech_committed', userMsg);\n this.#logger.child({ transcription }).debug('committed user speech');\n });\n\n this.#session.on('input_speech_started', (ev: any) => {\n this.emit('user_started_speaking');\n if (this.#playingHandle && !this.#playingHandle.done) {\n this.#playingHandle.interrupt();\n\n this.#session!.conversation.item.truncate(\n this.#playingHandle.itemId,\n this.#playingHandle.contentIndex,\n Math.floor((this.#playingHandle.audioSamples / 24000) * 1000),\n );\n\n this.#playingHandle = undefined;\n }\n\n const participantIdentity = this.linkedParticipant?.identity;\n const trackSid = this.subscribedTrack?.sid;\n if (participantIdentity && trackSid) {\n this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);\n }\n });\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n this.#session.on('input_speech_stopped', (ev: any) => {\n this.emit('user_stopped_speaking');\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_started', (ev: any) => {\n this.#pendingFunctionCalls.add(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_completed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n // eslint-disable-next-line @typescript-eslint/no-explicit-any\n this.#session.on('function_call_failed', (ev: any) => {\n this.#pendingFunctionCalls.delete(ev.callId);\n this.#updateState();\n });\n\n this.#session.on('metrics_collected', (metrics: MultimodalLLMMetrics) => {\n this.emit('metrics_collected', metrics);\n });\n\n resolve(this.#session);\n });\n }\n\n #linkParticipant(participantIdentity: string): void {\n if (!this.room) {\n this.#logger.error('Room is not set');\n return;\n }\n\n this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;\n if (!this.linkedParticipant) {\n this.#logger.error(`Participant with identity ${participantIdentity} not found`);\n return;\n }\n\n if (this.linkedParticipant.trackPublications.size > 0) {\n this.#subscribeToMicrophone();\n }\n\n // also check if already subscribed\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {\n this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);\n break;\n }\n }\n }\n\n #subscribeToMicrophone(): void {\n if (!this.linkedParticipant) {\n this.#logger.error('Participant is not set');\n return;\n }\n\n let microphonePublication: RemoteTrackPublication | undefined = undefined;\n for (const publication of this.linkedParticipant.trackPublications.values()) {\n if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n microphonePublication = publication;\n break;\n }\n }\n if (!microphonePublication) {\n return;\n }\n\n if (!microphonePublication.subscribed) {\n microphonePublication.setSubscribed(true);\n }\n }\n\n #handleTrackSubscription(\n track: RemoteTrack,\n publication: RemoteTrackPublication,\n participant: RemoteParticipant,\n ) {\n if (\n publication.source !== TrackSource.SOURCE_MICROPHONE ||\n participant.identity !== this.linkedParticipant?.identity\n ) {\n return;\n }\n const readAudioStreamTask = async (audioStream: AudioStream) => {\n const bstream = new AudioByteStream(\n this.model.sampleRate,\n this.model.numChannels,\n this.model.inFrameSize,\n );\n\n for await (const frame of audioStream) {\n const audioData = frame.data;\n for (const frame of bstream.write(audioData.buffer)) {\n this.#session!.inputAudioBuffer.append(frame);\n }\n }\n };\n this.subscribedTrack = track;\n\n this.readMicroTask = new Promise<void>((resolve, reject) => {\n readAudioStreamTask(new AudioStream(track, this.model.sampleRate, this.model.numChannels))\n .then(resolve)\n .catch(reject);\n });\n }\n\n #getLocalTrackSid(): string | null {\n if (!this.#localTrackSid && this.room && this.room.localParticipant) {\n this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant!.identity!);\n }\n return this.#localTrackSid;\n }\n\n #publishTranscription(\n participantIdentity: string,\n trackSid: string,\n text: string,\n isFinal: boolean,\n id: string,\n ): void {\n this.#logger.debug(\n `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`,\n );\n if (!this.room?.localParticipant) {\n this.#logger.error('Room or local participant not set');\n return;\n }\n\n this.room.localParticipant.publishTranscription({\n participantIdentity,\n trackSid,\n segments: [\n {\n text,\n final: isFinal,\n id,\n startTime: BigInt(0),\n endTime: BigInt(0),\n language: '',\n },\n ],\n });\n }\n\n #updateState() {\n let newState: AgentState = 'initializing';\n if (this.#pendingFunctionCalls.size > 0) {\n newState = 'thinking';\n } else if (this.#speaking) {\n newState = 'speaking';\n } else if (this.#started) {\n newState = 'listening';\n }\n\n this.#setState(newState);\n }\n\n #setState(state: AgentState) {\n if (this.room?.isConnected && this.room.localParticipant) {\n const currentState = this.room.localParticipant.attributes![AGENT_STATE_ATTRIBUTE];\n if (currentState !== state) {\n this.room.localParticipant.setAttributes({\n [AGENT_STATE_ATTRIBUTE]: state,\n });\n this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWA,sBAOO;AACP,yBAA6B;AAC7B,mBAAgC;AAChC,UAAqB;AACrB,iBAAoB;AAEpB,2BAA8D;AAC9D,mBAAiC;AACjC,2BAAiD;AAM1C,MAAe,wBAAwB,gCAAa;AAO3D;AAMO,MAAe,cAAc;AAQpC;AAGO,MAAM,wBAAwB;AAG9B,MAAM,wBAAwB,gCAAa;AAAA,EAChD;AAAA,EACA,OAAoB;AAAA,EACpB,oBAA8C;AAAA,EAC9C,kBAA2C;AAAA,EAC3C,gBAAsC;AAAA,EAEtC,uBAAuB;AAAA,EACvB;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA,yBAAyB;AAAA,EAC3B,GAKG;AACD,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAU;AACf,SAAK,0BAA0B;AAAA,EACjC;AAAA,EAEA,eAAkD;AAAA,EAClD,oBAAkD;AAAA,EAClD,iBAAgC;AAAA,EAChC,eAAmC;AAAA,EACnC,gBAAqC;AAAA,EACrC,iBAA4C;AAAA,EAC5C,cAAU,gBAAI;AAAA,EACd,WAAmC;AAAA,EACnC,UAA2C;AAAA,EAC3C,WAAwC;AAAA,EAExC,YAAqB;AAAA,EACrB,yBAAsC,oBAAI,IAAI;AAAA,EAC9C,aAAsB;AAAA,EAEtB,IAAI,SAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,OAAO,KAAsC;AAC/C,SAAK,UAAU;AACf,QAAI,KAAK,UAAU;AACjB,WAAK,SAAS,SAAS;AAAA,IACzB;AAAA,EACF;AAAA,EAEA,IAAI,wBAAqC;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAAsB,OAAoB;AAC5C,SAAK,yBAAyB;AAC9B,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,YAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAU,YAAqB;AACjC,SAAK,aAAa;AAClB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAS,SAAkB;AAC7B,SAAK,YAAY;AACjB,SAAK,aAAa;AAAA,EACpB;AAAA,EAEA,MACE,MACA,cAAiD,MACvB;AAC1B,WAAO,IAAI,QAAQ,OAAO,SAAS,WAAW;AAhJlD;AAiJM,UAAI,KAAK,UAAU;AACjB,eAAO,IAAI,MAAM,iCAAiC,CAAC;AAAA,MACrD;AACA,WAAK,aAAa;AAElB,WAAK,GAAG,0BAAU,sBAAsB,CAACA,iBAAmC;AAE1E,YAAI,KAAK,mBAAmB;AAC1B;AAAA,QACF;AACA,aAAK,iBAAiBA,aAAY,QAAS;AAAA,MAC7C,CAAC;AACD,WAAK;AAAA,QACH,0BAAU;AAAA,QACV,CAAC,kBAA0CA,iBAAmC;AAC5E,cACE,KAAK,qBACLA,aAAY,aAAa,KAAK,kBAAkB,YAChD,iBAAiB,WAAW,4BAAY,qBACxC,CAAC,iBAAiB,YAClB;AACA,6BAAiB,cAAc,IAAI;AAAA,UACrC;AAAA,QACF;AAAA,MACF;AACA,WAAK,GAAG,0BAAU,iBAAiB,KAAK,yBAAyB,KAAK,IAAI,CAAC;AAE3E,WAAK,OAAO;AACZ,WAAK,eAAe;AAEpB,WAAK,eAAe,IAAI,4BAAY,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW;AACjF,WAAK,gBAAgB,IAAI;AAAA,QACvB,KAAK;AAAA,QACL,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AACA,YAAM,mBAAmB,MAAM;AAC7B,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AAAA,MACnB;AAEA,YAAM,mBAAmB,CAAC,gBAAyB;AACjD,aAAK,KAAK,wBAAwB;AAClC,aAAK,YAAY;AACjB,YAAI,KAAK,gBAAgB;AACvB,cAAI,OAAO,KAAK,eAAe,aAAa;AAC5C,cAAI,aAAa;AACf,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,YAAY,OAAO;AAAA,YACjC,MAAM,IAAI,SAAS;AAAA,YACnB;AAAA,UACF,CAAC;AAED,cAAI,aAAa;AACf,iBAAK,KAAK,4BAA4B,GAAG;AAAA,UAC3C,OAAO;AACL,iBAAK,KAAK,0BAA0B,GAAG;AAAA,UACzC;AACA,eAAK,QAAQ,MAAM,EAAE,eAAe,MAAM,YAAY,CAAC,EAAE,MAAM,wBAAwB;AAAA,QACzF;AAAA,MACF;AAEA,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AACzD,WAAK,cAAc,GAAG,mBAAmB,gBAAgB;AAEzD,YAAM,QAAQ,gCAAgB,iBAAiB,mBAAmB,KAAK,YAAY;AACnF,YAAM,UAAU,IAAI,oCAAoB;AACxC,cAAQ,SAAS,4BAAY;AAC7B,WAAK,oBAAqB,QAAM,UAAK,qBAAL,mBAAuB,aAAa,OAAO,aAAa;AACxF,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,QAAQ,MAAM,yBAAyB;AAC5C,eAAO,IAAI,MAAM,yBAAyB,CAAC;AAC3C;AAAA,MACF;AAEA,YAAM,KAAK,kBAAkB,oBAAoB;AAEjD,UAAI,aAAa;AACf,YAAI,OAAO,gBAAgB,UAAU;AACnC,eAAK,iBAAiB,WAAW;AAAA,QACnC,OAAO;AACL,eAAK,iBAAiB,YAAY,QAAS;AAAA,QAC7C;AAAA,MACF,OAAO;AAEL,mBAAWA,gBAAe,KAAK,mBAAmB,OAAO,GAAG;AAC1D,eAAK,iBAAiBA,aAAY,QAAS;AAC3C;AAAA,QACF;AAAA,MACF;AAEA,WAAK,WAAW,KAAK,MAAM,QAAQ,EAAE,QAAQ,KAAK,SAAS,SAAS,KAAK,SAAS,CAAC;AACnF,WAAK,WAAW;AAGhB,WAAK,SAAS,GAAG,0BAA0B,CAAC,YAAiB;AAnPnE,YAAAC;AAqPQ,YAAI,QAAQ,gBAAgB,OAAQ;AAEpC,cAAM,eAAe,IAAI,2CAAsB,2CAAsB;AACrE,qBAAa,GAAG,eAAe,CAAC,SAAS;AACvC,eAAK;AAAA,YACH,KAAK,KAAM,iBAAkB;AAAA,YAC7B,KAAK,kBAAkB;AAAA,YACvB,KAAK;AAAA,YACL,KAAK;AAAA,YACL,KAAK;AAAA,UACP;AAAA,QACF,CAAC;AAED,cAAM,UAASA,MAAA,KAAK,kBAAL,gBAAAA,IAAoB;AAAA,UACjC,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR;AAAA,UACA,QAAQ;AAAA,UACR,QAAQ;AAAA;AAEV,aAAK,iBAAiB;AAAA,MACxB,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,YAAiB;AAE1D,YAAI,QAAQ,gBAAgB,QAAQ;AAClC,cAAI,KAAK,wBAAwB,KAAK,yBAAyB;AAC7D,kBAAM,IAAI;AAAA,cACR,0DACW,KAAK,uBAAuB;AAAA,YAGzC;AAAA,UACF;AAEA,eAAK;AACL,eAAK,QACF,MAAM;AAAA,YACL,QAAQ,QAAQ;AAAA,YAChB,MAAM,QAAQ;AAAA,YACd,SAAS,KAAK;AAAA,UAChB,CAAC,EACA;AAAA,YACC;AAAA,UAEF;AACF,eAAK,SAAU,wBAAwB,QAAQ,MAAM;AAAA,QACvD,OAAO;AACL,eAAK,uBAAuB;AAAA,QAC9B;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,0BAA0B,CAAC,OAAY;AA3S9D,YAAAA,KAAA;AA6SQ,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wCAAwC,CAAC,OAAY;AAvT5E,YAAAA,KAAA;AAyTQ,cAAM,gBAAgB,GAAG;AACzB,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,eAAe,MAAM,GAAG,MAAM;AAAA,QAC1F,OAAO;AACL,eAAK,QAAQ,MAAM,8BAA8B;AAAA,QACnD;AACA,cAAM,UAAU,IAAI,YAAY,OAAO;AAAA,UACrC,MAAM,IAAI,SAAS;AAAA,UACnB,MAAM;AAAA,QACR,CAAC;AACD,aAAK,KAAK,yBAAyB,OAAO;AAC1C,aAAK,QAAQ,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM,uBAAuB;AAAA,MACrE,CAAC;AAED,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AAzU5D,YAAAA,KAAA;AA0UQ,aAAK,KAAK,uBAAuB;AACjC,YAAI,KAAK,kBAAkB,CAAC,KAAK,eAAe,MAAM;AACpD,eAAK,eAAe,UAAU;AAE9B,eAAK,SAAU,aAAa,KAAK;AAAA,YAC/B,KAAK,eAAe;AAAA,YACpB,KAAK,eAAe;AAAA,YACpB,KAAK,MAAO,KAAK,eAAe,eAAe,OAAS,GAAI;AAAA,UAC9D;AAEA,eAAK,iBAAiB;AAAA,QACxB;AAEA,cAAM,uBAAsBA,MAAA,KAAK,sBAAL,gBAAAA,IAAwB;AACpD,cAAM,YAAW,UAAK,oBAAL,mBAAsB;AACvC,YAAI,uBAAuB,UAAU;AACnC,eAAK,sBAAsB,qBAAqB,UAAU,UAAK,OAAO,GAAG,MAAM;AAAA,QACjF;AAAA,MACF,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,KAAK,uBAAuB;AAAA,MACnC,CAAC;AAGD,WAAK,SAAS,GAAG,yBAAyB,CAAC,OAAY;AACrD,aAAK,sBAAsB,IAAI,GAAG,MAAM;AACxC,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,2BAA2B,CAAC,OAAY;AACvD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAGD,WAAK,SAAS,GAAG,wBAAwB,CAAC,OAAY;AACpD,aAAK,sBAAsB,OAAO,GAAG,MAAM;AAC3C,aAAK,aAAa;AAAA,MACpB,CAAC;AAED,WAAK,SAAS,GAAG,qBAAqB,CAAC,YAAkC;AACvE,aAAK,KAAK,qBAAqB,OAAO;AAAA,MACxC,CAAC;AAED,cAAQ,KAAK,QAAQ;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,qBAAmC;AAClD,QAAI,CAAC,KAAK,MAAM;AACd,WAAK,QAAQ,MAAM,iBAAiB;AACpC;AAAA,IACF;AAEA,SAAK,oBAAoB,KAAK,KAAK,mBAAmB,IAAI,mBAAmB,KAAK;AAClF,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,6BAA6B,mBAAmB,YAAY;AAC/E;AAAA,IACF;AAEA,QAAI,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AACrD,WAAK,uBAAuB;AAAA,IAC9B;AAGA,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,4BAAY,qBAAqB,YAAY,OAAO;AAC7E,aAAK,yBAAyB,YAAY,OAAO,aAAa,KAAK,iBAAiB;AACpF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,kBAAkB,kBAAkB,OAAO,GAAG;AAC3E,UAAI,YAAY,WAAW,4BAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAAA,EACF;AAAA,EAEA,yBACE,OACA,aACA,aACA;AAhbJ;AAibI,QACE,YAAY,WAAW,4BAAY,qBACnC,YAAY,eAAa,UAAK,sBAAL,mBAAwB,WACjD;AACA;AAAA,IACF;AACA,UAAM,sBAAsB,OAAO,gBAA6B;AAC9D,YAAM,UAAU,IAAI;AAAA,QAClB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,MACb;AAEA,uBAAiB,SAAS,aAAa;AACrC,cAAM,YAAY,MAAM;AACxB,mBAAWC,UAAS,QAAQ,MAAM,UAAU,MAAM,GAAG;AACnD,eAAK,SAAU,iBAAiB,OAAOA,MAAK;AAAA,QAC9C;AAAA,MACF;AAAA,IACF;AACA,SAAK,kBAAkB;AAEvB,SAAK,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC1D,0BAAoB,IAAI,4BAAY,OAAO,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW,CAAC,EACtF,KAAK,OAAO,EACZ,MAAM,MAAM;AAAA,IACjB,CAAC;AAAA,EACH;AAAA,EAEA,oBAAmC;AACjC,QAAI,CAAC,KAAK,kBAAkB,KAAK,QAAQ,KAAK,KAAK,kBAAkB;AACnE,WAAK,qBAAiB,+BAAiB,KAAK,MAAM,KAAK,KAAK,iBAAkB,QAAS;AAAA,IACzF;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,sBACE,qBACA,UACA,MACA,SACA,IACM;AA3dV;AA4dI,SAAK,QAAQ;AAAA,MACX,4BAA4B,mBAAmB,IAAI,QAAQ,IAAI,IAAI,IAAI,OAAO,IAAI,EAAE;AAAA,IACtF;AACA,QAAI,GAAC,UAAK,SAAL,mBAAW,mBAAkB;AAChC,WAAK,QAAQ,MAAM,mCAAmC;AACtD;AAAA,IACF;AAEA,SAAK,KAAK,iBAAiB,qBAAqB;AAAA,MAC9C;AAAA,MACA;AAAA,MACA,UAAU;AAAA,QACR;AAAA,UACE;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,WAAW,OAAO,CAAC;AAAA,UACnB,SAAS,OAAO,CAAC;AAAA,UACjB,UAAU;AAAA,QACZ;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,eAAe;AACb,QAAI,WAAuB;AAC3B,QAAI,KAAK,sBAAsB,OAAO,GAAG;AACvC,iBAAW;AAAA,IACb,WAAW,KAAK,WAAW;AACzB,iBAAW;AAAA,IACb,WAAW,KAAK,UAAU;AACxB,iBAAW;AAAA,IACb;AAEA,SAAK,UAAU,QAAQ;AAAA,EACzB;AAAA,EAEA,UAAU,OAAmB;AAjgB/B;AAkgBI,UAAI,UAAK,SAAL,mBAAW,gBAAe,KAAK,KAAK,kBAAkB;AACxD,YAAM,eAAe,KAAK,KAAK,iBAAiB,WAAY,qBAAqB;AACjF,UAAI,iBAAiB,OAAO;AAC1B,aAAK,KAAK,iBAAiB,cAAc;AAAA,UACvC,CAAC,qBAAqB,GAAG;AAAA,QAC3B,CAAC;AACD,aAAK,QAAQ,MAAM,GAAG,qBAAqB,KAAK,YAAY,MAAM,KAAK,EAAE;AAAA,MAC3E;AAAA,IACF;AAAA,EACF;AACF;","names":["participant","_a","frame"]}
@@ -1 +1 @@
1
- {"version":3,"file":"multimodal_agent.d.ts","sourceRoot":"","sources":["../../src/multimodal/multimodal_agent.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAEV,gBAAgB,EAChB,iBAAiB,EAGjB,IAAI,EACL,MAAM,mBAAmB,CAAC;AAS3B,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAE3C,OAAO,KAAK,GAAG,MAAM,iBAAiB,CAAC;AAOvC;;;GAGG;AACH,8BAAsB,eAAgB,SAAQ,YAAY;IAExD,QAAQ,CAAC,YAAY,EAAE,GAAG,CAAC;IAE3B,QAAQ,CAAC,gBAAgB,EAAE,GAAG,CAAC;IAC/B,QAAQ,CAAC,MAAM,EAAE,GAAG,CAAC,eAAe,GAAG,SAAS,CAAC;IACjD,QAAQ,CAAC,uBAAuB,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI;CACvD;AAED;;;GAGG;AACH,8BAAsB,aAAa;IAEjC,QAAQ,CAAC,OAAO,CAAC,OAAO,EAAE,GAAG,GAAG,eAAe;IAC/C,QAAQ,CAAC,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAC/B,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;CAC/B;AAED,MAAM,MAAM,UAAU,GAAG,cAAc,GAAG,UAAU,GAAG,WAAW,GAAG,UAAU,CAAC;AAChF,eAAO,MAAM,qBAAqB,mBAAmB,CAAC;AAEtD,YAAY;AACZ,qBAAa,eAAgB,SAAQ,YAAY;;IAC/C,KAAK,EAAE,aAAa,CAAC;IACrB,IAAI,EAAE,IAAI,GAAG,IAAI,CAAQ;IACzB,iBAAiB,EAAE,iBAAiB,GAAG,IAAI,CAAQ;IACnD,eAAe,EAAE,gBAAgB,GAAG,IAAI,CAAQ;IAChD,aAAa,EAAE,OAAO,CAAC,IAAI,CAAC,GAAG,IAAI,CAAQ;gBAK/B,EACV,KAAK,EACL,OAAO,EACP,MAAM,EACN,sBAA0B,GAC3B,EAAE;QACD,KAAK,EAAE,aAAa,CAAC;QACrB,OAAO,CAAC,EAAE,GAAG,CAAC,WAAW,CAAC;QAC1B,MAAM,CAAC,EAAE,GAAG,CAAC,eAAe,CAAC;QAC7B,sBAAsB,CAAC,EAAE,MAAM,CAAC;KACjC;IAuBD,IAAI,MAAM,IAAI,GAAG,CAAC,eAAe,GAAG,SAAS,CAE5C;IAED,IAAI,MAAM,CAAC,GAAG,EAAE,GAAG,CAAC,eAAe,GAAG,SAAS,EAK9C;IA6BD,KAAK,CACH,IAAI,EAAE,IAAI,EACV,WAAW,GAAE,iBAAiB,GAAG,MAAM,GAAG,IAAW,GACpD,OAAO,CAAC,eAAe,CAAC;CAyX5B"}
1
+ {"version":3,"file":"multimodal_agent.d.ts","sourceRoot":"","sources":["../../src/multimodal/multimodal_agent.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAEV,gBAAgB,EAChB,iBAAiB,EAGjB,IAAI,EACL,MAAM,mBAAmB,CAAC;AAS3B,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAE3C,OAAO,KAAK,GAAG,MAAM,iBAAiB,CAAC;AAOvC;;;GAGG;AACH,8BAAsB,eAAgB,SAAQ,YAAY;IAExD,QAAQ,CAAC,YAAY,EAAE,GAAG,CAAC;IAE3B,QAAQ,CAAC,gBAAgB,EAAE,GAAG,CAAC;IAC/B,QAAQ,CAAC,MAAM,EAAE,GAAG,CAAC,eAAe,GAAG,SAAS,CAAC;IACjD,QAAQ,CAAC,uBAAuB,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI;CACvD;AAED;;;GAGG;AACH,8BAAsB,aAAa;IAEjC,QAAQ,CAAC,OAAO,CAAC,OAAO,EAAE,GAAG,GAAG,eAAe;IAC/C,QAAQ,CAAC,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAC/B,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;CAC/B;AAED,MAAM,MAAM,UAAU,GAAG,cAAc,GAAG,UAAU,GAAG,WAAW,GAAG,UAAU,CAAC;AAChF,eAAO,MAAM,qBAAqB,mBAAmB,CAAC;AAEtD,YAAY;AACZ,qBAAa,eAAgB,SAAQ,YAAY;;IAC/C,KAAK,EAAE,aAAa,CAAC;IACrB,IAAI,EAAE,IAAI,GAAG,IAAI,CAAQ;IACzB,iBAAiB,EAAE,iBAAiB,GAAG,IAAI,CAAQ;IACnD,eAAe,EAAE,gBAAgB,GAAG,IAAI,CAAQ;IAChD,aAAa,EAAE,OAAO,CAAC,IAAI,CAAC,GAAG,IAAI,CAAQ;gBAK/B,EACV,KAAK,EACL,OAAO,EACP,MAAM,EACN,sBAA0B,GAC3B,EAAE;QACD,KAAK,EAAE,aAAa,CAAC;QACrB,OAAO,CAAC,EAAE,GAAG,CAAC,WAAW,CAAC;QAC1B,MAAM,CAAC,EAAE,GAAG,CAAC,eAAe,CAAC;QAC7B,sBAAsB,CAAC,EAAE,MAAM,CAAC;KACjC;IAuBD,IAAI,MAAM,IAAI,GAAG,CAAC,eAAe,GAAG,SAAS,CAE5C;IAED,IAAI,MAAM,CAAC,GAAG,EAAE,GAAG,CAAC,eAAe,GAAG,SAAS,EAK9C;IA6BD,KAAK,CACH,IAAI,EAAE,IAAI,EACV,WAAW,GAAE,iBAAiB,GAAG,MAAM,GAAG,IAAW,GACpD,OAAO,CAAC,eAAe,CAAC;CA6X5B"}
@@ -10,7 +10,7 @@ import { EventEmitter } from "node:events";
10
10
  import { AudioByteStream } from "../audio.js";
11
11
  import * as llm from "../llm/index.js";
12
12
  import { log } from "../log.js";
13
- import { BasicTranscriptionForwarder } from "../transcription.js";
13
+ import { TextAudioSynchronizer, defaultTextSyncOptions } from "../transcription.js";
14
14
  import { findMicroTrackId } from "../utils.js";
15
15
  import { AgentPlayout } from "./agent_playout.js";
16
16
  class RealtimeSession extends EventEmitter {
@@ -121,7 +121,7 @@ class MultimodalAgent extends EventEmitter {
121
121
  this.emit("agent_stopped_speaking");
122
122
  this.#speaking = false;
123
123
  if (this.#playingHandle) {
124
- let text = this.#playingHandle.transcriptionFwd.text;
124
+ let text = this.#playingHandle.synchronizer.playedText;
125
125
  if (interrupted) {
126
126
  text += "\u2026";
127
127
  }
@@ -166,16 +166,20 @@ class MultimodalAgent extends EventEmitter {
166
166
  this.#session.on("response_content_added", (message) => {
167
167
  var _a2;
168
168
  if (message.contentType === "text") return;
169
- const trFwd = new BasicTranscriptionForwarder(
170
- this.room,
171
- this.room.localParticipant.identity,
172
- this.#getLocalTrackSid(),
173
- message.responseId
174
- );
169
+ const synchronizer = new TextAudioSynchronizer(defaultTextSyncOptions);
170
+ synchronizer.on("textUpdated", (text) => {
171
+ this.#publishTranscription(
172
+ this.room.localParticipant.identity,
173
+ this.#getLocalTrackSid(),
174
+ text.text,
175
+ text.final,
176
+ text.id
177
+ );
178
+ });
175
179
  const handle = (_a2 = this.#agentPlayout) == null ? void 0 : _a2.play(
176
180
  message.itemId,
177
181
  message.contentIndex,
178
- trFwd,
182
+ synchronizer,
179
183
  message.textStream,
180
184
  message.audioStream
181
185
  );